The Oxford Internet Policy & Politics Hybrid Conference 2024

This stream is for Room 1.

For more information, streaming links to the other rooms, and the full program please visit: https://ipsonet.org/conferences/ippconference/

E can we move on to the next slide please right thank you um and uh of course um the there’s been uh more research on the topic of uh social fragmentation and polarization um not only recently but a few years ago uh we can see here for example some of

The reports published here in the United Kingdom uh by uh the policy Institute at King’s College London um in September 2019 um this study was launched uh following up the uh brexit vote in the United Kingdom um on the right here uh there is a another study by pure

Research Center in the United States uh which uh looked at political polarization in the American uh public um and of course there were some uh interesting conclusions in in those studies and I think quite notable is the uh study of the PE Research Center um where they found how increasing

Ideological inity and partisan antipathy can affect politics um compromise and everyday life and uh of course to many of us maybe today is not a surprise but at that time um the and media content regulation particularly in the early stages of the pandemic governments and their advisers were dealing with significant

Uncertainty conflicting information and advice circulating on the internet added to the confusion and fueled Elite Panic about misinformation and disinformation in an interview in March 2020 prime minister jinda ardun acknowledged that our understanding of covid-19 was evolving rapidly and that rumors were circulating on social media she committed to sharing the most

Upto-date information daily and on the ministry of Health’s website urging trust in government as the source of that information she went on to say we will continue to be your single source of truth that we’ll provide information frequently we will share everything we can everything else you see a grain of

Salt um he’s going to talk to us about reping in progress about um why do you online is that we wanted to study the role of such media in um social conflicts and uh that result from a group po polarization uh next slide please um thank you um yeah so uh of course

Looking uh for uh back on the role of media for democracy uh of course um the initial thought would be that the media U can contribute to pluralism in the opinions uh and uh it’s we can take this quote from Hamilton as his justification for um the positive uh uh influence of U

Democratic deliberation or the so the the debates in in the legislative branch that can promote um deliberation and circumspection and then of course it can serve as a check uh excesses and U uh to the excesses of the majority uh if we look at the literature next slide please

Uh if we look uh at earlyer theories of democratization uh of course uh this modernization Theory with its assumptions about the Democracy as a byproduct of 74 sharing services and 19 other means of communication you can see the the the general diagram visible on the left the the green brackets are

Those that were accessible uh the red ones were inaccessible but still identifiable and the white ones were not monitored because they were for instance encrypted communication apps or file sharing services but General generally speaking it was a quite large um uh information ecosystem maintained by the group that’s the primary target of of

Content takedown so it shouldn’t be that developed however the biggest surprise was uh made by the fact that uh in the second part of the project uh that was carried out six months later in June 2021 uh I I carried out the second round of of the investigation and and I’ve

Learned that the size of the home whole network propaganda dissemination Network almost tripled to uh 386 communication channels you can see again the diagram on the left and it consisted of 58 Standalone websites uh 12 social media profiles more than 285 sharing services and 34 uh other means of communication

Mostly these were um encrypted communication apps telegram frea um and also I will provide some examples um so so number one is the fragmenting and polarizing effect and uh as uh as it is often stated uh I think it was President Obama who said that uh uh person who

Reads New York Times lives in a different reality U compared to somebody who uh watches uh Fox News for example and uh so we it led to uh the the world where uh it’s now just borrowing a uh a popular term it’s not uh the world of imagined communities but the world of

Imagined or created realities uh and of course that leads to more misunderstanding and the increased risk of conflict next slide please uh so the group polarization uh would mean uh according to sunstein that members of deliberating group uh moved towards a more extreme point in the direction indicated by the the members predeliberation

Tendencies uh can we skip the next and then move move to the other yes next please uh so in in simple words can we go one more uh in simple words the group polarization uh is making decisions in contexts of uncertainty we won’t always get it right and we will make

Mistakes uh fourthly nurture and model a culture of free speech and open debate answer false or misleading information with accurate information and reasoned argument communicate rapidly accurately positively and honestly and finally um exposing shaming and ridiculing opponents does not work and will only increase polarization exercise and enable

Effective of counter speech that calls people in instead of calling them out be curious and interested in why people disagree these are the strategies I offer as alternatives to aite panic about Miss and disinformation and the urge to censor acknowledge pluralism practice and institutionalized Toleration desist from labeling criticism disagreement and descent as

Missile disinformation nurture and not accurate prediction next please um next uh one more please thank you and the and the examples are there are plenty examples both in the west uh as with Trump elections and the brexit in the United Kingdom but also with u the Russian invasion uh in Ukraine that is

Fueled by Propaganda and nationalistic sentiments and more recent U statements by Russian politicians about a few other issues uh related to the relations of Russia and Kazakhstan and in particular the quite often mentioned topic is the decline in the use of Russian language in Kazakhstan and the uh decline in the

Number of Russian language schools uh which um uh as we could see the national nationalistic sentiments are often used uh by politicians to uh gain political uh uh um points uh or to get the more support from the uh uh from the population next please and the final finally we have the

Stent effect uh the stent effect that that’s one of the uh side effects of of content takedowns the stent effect is based on on a perceived increased value of the sensored content on the internet um which forces some internet users to re-echo to share uh the content

That’s been taken down uh on the on the internet and unfortunately we could see similar reactions uh to to terrorist content being taken down because some internet users just dis like uh censorship on the internet and we could see it for instance in 2014 2015 when even totally unaffiliated internet users

That do not follow the the radical agenda of Islamic State uh they tended to to share um their propaganda just for the sake of fighting against censorship so about theist innovations that I mentioned a moment ago some of the most uh important ones in the last four five

Years include um the introduction of the anti crawling and antios techniques at Terrace operated websites which means that they are much harder to be taken down by cyber attacks for instance consequences was spreading really fast in in social media whereas the official channels uh remained uh uh

Largely silent uh until the next day or until till uh the next morning where people were already informed uh uh with both uh uh false or pulse news and some people had information about casualties which was turned out not to be true next please um of course the uh there is also

Need to to use um uh to to use interdisciplinary approach in in these studies and uh uh of course um uh it’s um it’s used widely to spread propaganda and using the public money quite often uh to create uh boat Farms um or uh to influence from outside or organizing

Cyber attacks and there were several uh cyber attacks reported recently on uh independent journalists for example in in our in the country where I come from and the surveillance is to some extent could be also made easier for government agencies because uh it would um um be

Would be uh um quite straightforward to sometimes to identify dissidents and uh and then follow with our repressions I think anything should be done about this whether you think it’s a problem um or whether whether you think that the live and live mentality even still even

Hold holds in an age where people can be ex extremely manipulated uh online thank you for your question um it it lets me clarify first that I am not a free speech absolutist or a free speech purist um I did mention briefly in my presentation uh that an international

Human rights law uh the right to freedom of speech is a qualified right um and states can justifiably prohibit uh the incitement of uh acts of hostility discrimination and violence uh so there are limits to the exercise of Free Speech but in terms of the the point you rais about um yes

There is false and misleading information circulating on the internet and some of it does real harm so the question is who do you trust way uh contribute to at least minimizing the nasties on the internet but that’s going to be a complex process and uh as someone who

Worked for nearly 20 years in government I do not have the confidence that uh either the governments are competent to deal with us um or that it would be a wise to give them the power to do so so it’s um I I’m I’m not saying that um

Free speech is an absolute right I’m just asking the question if we want to restrict the exercise of freedom of opinion and expression uh what would be the kind of rules and who would be the rul makers that we would trust to do that on a responsible

Mayor terrific um I think we have time for maybe one more question if that’s all right with the professor we could fit one more in if anyone has one all right we so the will you sat down the table yeah’s oh oh I see if we it’s easier if we stand at the

Leon but Debbie will need a microphone cuz she want to stay sat down won’t she and we probably we probably won’t go jumping up and down if there are questions not well we can this my C up here yes so laurren thank you for coming we’ll be starting in minutes hello for us

Today thank you and good morning everyone or good evening to to to Nadia um let me start briefly with with the the aim of this session and how it fits with the with the general conference theme uh the conference team look or the conference looks at the implications the various diverse

Implications of the internet and related Technologies and since we are researchers and and analytics working in the field of working life we try to connect these two topics so we are looking at the consequences of internet and Technologies for working life and for labor market in in general and

Therefore the current session will uh look at or Zoom more closely at the different regulatory models uh that are out there for the gig economy from a Global Perspective uh we will also look and evaluate the challenges and opportunities that are there uh with the that the regulation of the gig economy

Is facing and we will explore the feasibility of collect secondly there are very real dangers that are associated with regulation being over prescriptive as it can you know stifle Innovation and third thirdly a practical implication of regulations is that they increase the cost of compliance and this increase in the cost

Of compliance can make the already uneven playing field a lot more unequal especially for innovators startups and smmes looking to develop Innovative breakthrough uh you know applications in Ai and how do we solve this we solve this by baking in sociotechnical interventions right from the outset that measure Monitor and mitigate a model’s

Risks throughout it life cycle we do this by leveraging two sorts of interventions of course there are many but I’ll cover two today model evaluations Cally called evals and audits uh while evals helped sort of gauge the model efficacy the the rate of model efficacy and mitigate risk during

A model’s life cycle audits are a powerful way of independently stress testing interrogating models against different sorts of risks um audits can also help assure models and can also act as very strong sign of Regulatory Compliance um and of course these are made poent and these can be made poent

You glimpse of how the platform themselves uh the trade unions the courts and the governments are doing it so the first is self-regulation by the platforms uh there is a crowdsourcing code of conduct it started in 2015 it has 10 principles such as Fair payment only serious tasks and open

And transparent communication 10 platforms have already signed the code of conduct in 2017 there were there was established and an office to effectively effectively implement the code of conduct and resolve any disputes between workers and signatory Platforms in 2020 the world economic Forum uh chter of principles for a good

Platform work it uh was signed by the major reg labor platform platforms including cabify Uber delivery and grab the charter commits the platforms to diversity and inclusion safety and well-being flexibility and fair conditions reasonable pay and fees and social protection and so many other issues then there is globo cous pledge

Globo introduced a cous pledge in 2021 it is being implemented in 17 countries today the the pled rests on four pillars and the most important of those is fair ings and safety and there’s Oh um and these can be developed internally uh or taken from public academic uh databases which is usually the case um model evaluations can be used throughout a models development life cycle and are used for health CL give me one second okay I’m assuming everyone can see um title screen here and everyone

Can hear me okay uh if not happy to speak louder or make whatever adjustments we need in order for you to be able to see the screen um I appreciate everyone um coming to listen a little bit to what I’ve got about to say um my talk is about the application of Behavioral

Cyber security and instant response um I think that um what I want to try to talk about today is um the application of some of the elements that um we we see and we find in behavioral economics um in the world of cyber security and as I I kind of move through the

Presentation I going to give you a couple a short rundown on um some basic economic Concepts a little bit about behavioral economics and then how those um key elements from behavioral economics really are are quite applicable in the cyber security realm and then talk about um inant response

And root cause analysis and how we can kind of apply those to come up with perhaps a better or more meaningful way of looking at cyber security and cyber security and response um the efficacy of it reducing the prevalence of violence and hate sexual content criminal planning guns and

Illegal weapons self harm regulated substances against you know sort of competing models like open AI mod API perspective API using benchmarking techniques this is just an example of how benchmarking kind of you know plays out however benchmarks are well not perfect and have limitations firstly there’s a perceptible lack of appropriate benchmarks

Across model harms and modalities I’ll kind of get into net benefit or the most positive outcome for their um for their own personal behavior and if we think about you know tying rational actor and rational agent from behavioral economics or regular economics um we we know that people don’t necessarily always do any

Of these we don’t operate under perfect information we we sometimes I have inconsistent preferences um we don’t always do cost benefit analysis and um therefore you know it was it was relatively easy to understand um and and see why um we we kind of needed to move

Towards a little bit different model as it related to just some basic um economic decision making and those kinds of things so out of that kind of raised or arose this this concept of Behavioral economics which was more of a multi-disciplinary approach that offers some a little bit more precise and

Comprehensive understanding of economic activities and Market phenomenon and um it it acknowledges that individual decision making is is really constrained by their cognitive limitations particularly um you know not necessarily having the most perfect information um to them or available to them all the time uh and that people process

Information differently um we come to decision making with biases and um mental shortcuts and and those types of things that lead to quite often irrational decisions um which is kind of contrary to the traditional behavioral economics model and and I think the really neat thing here is is that we can

Apply a lot of these things to cyber security and cyber security policymaking that we see both now and hopefully in the future where a lot of policy making within organizations is based upon this Theory or this concept of a rational actor and that that individuals uh whether working

For the organization as a user with a non-cyber security role or those individuals that have a sa framework to help people with big IQs and smaller IQs effectively navigate through a situation number three it provides a meeting place for ideas and even taboo ideas when I talk to people in ideologically extreme

Positions If I argue with them we just get what resists persists we just fight we get stronger muscular if you like in our in our debate but if I say let me understand your ideology here is a counter ideology or here is another ideology or even within your ideology

Pull out the bits that actually are good that are useful rather than just everything is bad and useless so what I want is pull out those ideologic IAL positions that I think actually within that I can hear where you’re coming from I can see what you’re

Saying I I can kind of get my hands around what you’re trying to Grapple with because it resonates now how about if we situate that within this ideology how does that work and actually I find that people are far more open and receptive a to thinking in a different

Way but also if we want to get all security and psychological about it it actually is quite subversive because we’re helping change the narrative the third area the the fourth area is it provides a road map for policymaking so I would suggest that when it comes to

Education when it comes to um how how we how we bring our children up I think rather than just blanking what’s going on my you know my children my my sons now have been through school Etc and they were being confronted with the Andrew Tes of this world Etc now you can

Argue my one naughty word for the say you can argue the toss with people like that and I’ve sat with individuals who hold to those things and you can argue Point by Point you’re going to lose it’s not going to work but if we grab a hold

Of and say actually I can I think I can see where you’re coming from and then policy will be to equip our Educators and our education system with ways to train our children and our young people How to Think Through ideological lenses rather than just try to beat the other

Person through a sound argument and reason and actually rather than resist them actually to pull out that’s an interesting point where do we situate that Within These different ideological positions the Anthropologist Margaret me and Gregory Bateson looked at how societies change and evolve and they drew a simple pyramid a simple triangle

As all good models all good heuristics provider shortcuts and navigate complexity if we want to create change sometimes we can approach at the environmental level let’s change the chairs around you know physical level we can approach it at the skills level or the training level you need to

Understand this you need to do that Etc that’s a very transactional approach to creating change in society the fastest changes in society according to them and the research across the world is when you touch people at a belief or an ideological level because we are not just rational logical beings we are

Emotive meaning making human beings who need to Find meaning place and purpose in life so therefore an ideology Will trump an ideology Amid and Bateson and other people like that said if you get to the top of the pyramid which is around belief purpose and identity and

You start to converse around then the cascading change when people start to grab a hold of those positions is faster than if you try to drive transactional change from the bottom up so I would argue from a position of transformational language through ideology rather than transactional language through sheer raw argument in

And of itself doesn’t mean we shouldn’t be logical shouldn’t mean we shouldn’t mean we shouldn’t argue and bring facts in but we have to deal at the ideological level let me throw out a simple IDE ideological thought very small when I’m going to finish on this and it goes something like

This if we look at the world there’s enough enough for everybody’s need but not for everybody’s greed a simp simple ideological position a simple Paradigm which we can look at the world through now we can then start to dive into the complexities because how does that affect environment need and greed

How does that affect financial institutions need and greed how does that affect business need and greed how does that affect the marginalizing the elites of society need and greed how does that affect foreign policy and Aid expansionism need and greed simple py things like that make sense instinctively at an ideological level

And give us an opportunity to drop down and talk policy talk taboo have a framework and a moral compass of which to navigate complexity through it’s a massive challenge but we can’t ban It We Can’t Ban the stuff we see online it’s too late it’s got away it’s too

Democratized we need to engage get on the playing field and actually start to think up ideological positions that Inspire rather than seek to crush other people just for the sake of winning thank you very much I will be opening up for questions but I think we’re going to start with an

Interpanel discussion yes and I want to raise something which wasn’t raised Jonathan Hall KC who’s the um counter terrorism sort of regulator he said the internet is a major source of radicalization and self- radicalization of children and one of the problems with the Online safety Bill

He feels is that children are not being protected enough so could I have some thoughts on that and also um the head of the Mets um counterterrorism unit Commander Dominic Murphy has said that the me Met has seen an uncomfortable um rise in child radicalization driven by online content

And one of the issues is how do you protect children on the internet we know that Tik Tok um is huge even and when there even age restrictions that’s not necessarily um implemented so that follows on your point Paul about education and in Finland they have a a very effective I think educational

Policy on teaching children about recognizing misinformation disinformation so could we explore that a bit how do you feel that government you know could deal because it it and also in France they’ve been witnessing a huge uptick in is activity um recruiting in Isis recruiting a new generation of

Young people online and so could could we explore that a bit please thank you sure who wants to yeah um yes I’ve have more technical thoughts around this kind of stuff but I I engage both with um prevent and security around the S the Met side of things as well as

People on the ground and I think one of the things that we can do which is very very practical is from a social media perspective rather than try to to ban it and police and regulate which I think that would be lovely but I think would be on the stage

Where it’s even possible with how you can get around things these days that’s just the way it is I wonder whether we should start turning on the algorithms that actually search for facts I think there’s an algorithmic argument that says that a lot of the social media algorithms are Sensational um they pick

Up on certain Trends Etc that are more sensational well why don’t we disrupt that by turning on and releasing algorithms and chat Bots and machine learning and all the other bits and pieces that are actually uh fact finders truth Seekers um and do it that way um and the reason

Why I say that is because um I can’t mention names because we’re streaming but some people I was I was talking to some time ago worked for the parent company of one of the well-known um online social media platforms and one of one of his roles um um and he’s involved

With psychological warfare and he got employed by such an organization allegedly one of his roles was to um develop the algorithms that would persuade people to move from one position to another particularly when it came to the brexit argument and so I know it’s indeed possible for us to

Begin to work with social media companies to develop algorithms that actually pull out facts and truth and wrap them into some sort of clever ideological position uh it’s just using the same um methods that are used negatively but use them positively um thank you but so far a lot

Of the social media companies have been reticent they’re big business and one of the problems is sensationalism as we know um sales do you think I mean there’s a big argument that the internet companies should be held to account by not as being being held to account by the same standards um

That you know old media the newspapers that they should be treated as Publishers one of the big problems is that it’s quite difficult at the moment with current legis well the legislation just doesn’t keep up could you I’m the there’s a part of me that’s

A PR that’s quite pragmatic and I I I would love to think that we could turn back the clock and make make all those things happen I don’t know how you police it I don’t know how you regulate it effectively and efficiently for a long enough period of time um it’ll be

Fantastic it would be fantastic but so I’m I’m much more of the well why don’t we use the influence subversion uh approach subversion I don’t know if that’s the right word but the influence approach and somehow hook it to the bottom line of media companies as well

Or advertisers um and make make truth Sensational make accuracy Sensational exciting in some sort of more positive narrative but the it’s a carot anistic approach that you know if they’re held to account as Publishers some people feel that would improve you know standards because at the moment it seems

And obviously because these are crossborder international companies that that’s obviously another hurdle that you know who where what regulates I mean my my sense is that I not an expert on on policy and and law in that respect but I would think it’s nearly impossible to legislate unless unless the state steps

In and starts turning things off and clamping down and unfortunately what for my perspective at least I think that these companies and how shares work and how the stock exchange works and trans transnational um business operates I just don’t think Advanced Western democracies can operate in that

Way maybe it’s a bit sad I was going to say a little bit P William any thoughts any thoughts on this s i I think trying to get control the internet companies is it’s fine but it’s been done it’s being done it’s part of the process and it’s failing um the

Children uh the the the whole of society is subverting children not just just the internet you look at mainstream television what’s the big problem with children and that is violence knife crime anger aggression brutality that is creeping into their lives and it’s creeping into their lives from mainstream media from

The BBC and the ITV which put on programs which had frankly very very brutal now and very very brutal and and people become inured to violence desensitized it’s uh in my view this is my view I’m just saying and so what we need is a kind of sea change in society

And we also need if you want an oldfashioned view we’ve taken away all of the things that children did not just just in Britain but across the world because of this change in approach from from governments and our own government here in the UK you the youth clubs have gone the the uh

Swimming pools are being closed down the kinds of things that would occupy children and get them off the streets and you see a huge rise in knife crime you see a huge rise in violent crime and I think it’s a problem that we are not addressing properly and there is

Alienation these people these kids don’t feel they belong to anything they haven’t got a club to belong to they haven’t got an ideology to belong to um and what are the great ideologies they are socialism Isis moral rearmament um pacifism like Martin Luther King active P pacifism you know

You have all these kind of gripping thoughts well what is Isis it’s grasped a lot of young people and caused them to go and fight in Syria not I mean it’s a thing of the past but it may be a thing of tomorrow too especially with what’s

Going on in Gaza now when people are being radicalized by that uh what is what is Isis Isis is um an ideology that believes in the in Armageddon and exclusivity it believes that you’re with us or you’re against us and these are the final days how do we protect

Children how do you protect children from an ideology like this is by by promoting inclusivity and promoting uh an ideology basically of U you know you all belong you are part of our society and inclusivity is enormously important rather than the exclusive ideology that is being thrust down the inter internet

Whether you’re talking about the extreme right who hate anybody who aren’t the same as them or Isis who hate everybody who aren’t the same as them the you you need to fight this kind of extremism with inclusivity and with an ideology of love I’m I’m sorry it sounds

Terribly oldfashioned but it is I think the only way and it it has to be done and it has to be done as an alternative instead of merely trying to suppress extremism you will never win a battle where you just try and suppress extremism in my thank you William um I’d

Like to now open it up to questions from the audience um if you could introduce yourself and say where you’re from that would be great do you need yeah there’s one coming from the back okay there we go thank you hello yeah yeah is it on all right um

Thank you very much for the for the talk I’m going to kind of combine both talks in my question um my name is Tyler West and I’m doing a a PhD um in Ireland almost finished I’m looking at um user agency in media regulation social media regulation actually so this is this is

Quite uh pertinent but I’m also I’m from Appalachia and um I’m from I don’t think my community is necessarily unique in this regards I’ve seen so many people become maybe a soft version of radicalization um but also like you know conspiratorial thinking and it’s really disheartening to be honest

Um and so so your your talks resonated with me on that level um one of my questions was for uh when we’re talking about ideology and and meeting an ideology with another ideology how do you do that in a way first of all what does that alternative ideology look like

Um I’m assuming it isn’t necessarily a progressive ideology because that would be absolutely uh blocked from the get-go what does that look like and how do we keep that from being paternalistic because in some ways you’re coming in and you’re saying well I’m I’m taking what you’re saying on board but here I’m

Going to change it you know um so how do you make that work because in my experience when anybody I personally know I I try to challenge their ideologies or even act like I’m commiserating in some way with their beliefs they’re like I’m on to you um you’re

Trying to change my mind so that was one question the other really for both of you was um what is the cause of all this we talk about uh the the the anger and the anti- elite nobody really talked about what they thought the cause the root cause

Was and I think until we get to the root causes and fix those this problem will William touched on it thank you so if you could keep the answers brief yeah thank you um who wants to go first Paul do you want to go first Tyler thank you um yes I spent

Many many hours weeks months it feels like years in rooms talking to people with really strong ideologies and in the early stages I was a bit of a trickster I was thinking hey yeah how can I make you think that I I agree and then I can subvert because I got some clever

Socratic questioning technique doesn’t work actually the truth of the matter is this is this takes time and actually going in with two ears and one mouth using in those proportions and genuinely listening for actually I I can see where you’re coming from actually tell me more

Tell me more and it’s time and proximity that win people not the argument the arguments come later of course we can take people on a logical thing but actually it takes time to walk with people through a process how have I changed over time I’ve had Epiphany

Moments but it’s been over time whether I like it or not that’s just how we change brains like the shortest possible route they don’t like burning carbohydrates so they need time to change in terms of the causal factors really interesting whether I talk to criminal whistleblower through to

Pedophile Hunter the common thing that comes through is a sense of existential nothing nothingness it’s they call it black pilling what’s the point red pilling is hey we woken up to see the world as it really is but black pill is I’ve work up to se world as it really is

But there’s just no point and I think people are in an existential vacuum and I think that’s a big because we’re seeing we weren’t getting too geopolitical the collapsing of familiar things around us creates and I’m going to try and keep it py so I’m going to

Stop but I think there’s a lot to unpack in the existential vacuum and of course you get strong women and men stepping into that and providing Simple Solutions to complex answers and people want security and stability so they latch onto it and find meaning something to live for but more importantly something

To die for and I think we’ve kind of lost a lot of that something to die for business thank you William yeah I think that’s exactly right I mean the the problem you have is that people don’t have a sense of belonging enough the the old nationalisms have gone and they are

Replaced by a very parochial kind of attitude where you dislike the other and and that’s it I I don’t I don’t i’ never liked nationalism I’m an internationalist I believe in in a multicultural world but it’s not the there is a lack of a sense of

Belonging and I think that is a core problem with the alienated people don’t feel they belong and then hatred gets fostered and there’s um and hatred we live in a much more hating world now hatred is is much more of a feature of Our Lives um hatred of Israel hatred of

Of uh Russia it you you get a lot of hatred bouncing around amongst Ordinary People and it is it is a real problem that this is becoming a phenomena in our society in my view uh hatred of migrants hatred of the the hatred is becoming a um a major

Phenomena hatred of brexiteers by remainers and vice versa um so it’s it’s something we have to deal with but the other thing is ideologies for inclusivity I mean um we’ve experimented with this at my Foundation the one I work with um in different ways and one

One time few years back we set up an online group called Kata kata’s Arabic for Reflections and we decided we’d make it a kind of Sufi esque program and we built it on the basis of an organization Paul knows called moral rearmament which believes in or used to believe

In doesn’t exist now in uh really but used to believe in absolute Purity absolute truth absolute love and ab absolute selflessness now it’s it’s it was an odd thing to build an ideological page on we did it as an experiment and we made it totally inclusive So Pro gay

Pro everybody you know a a one world ideology of the kind of thing that Khalil Jil Brown Used To Believe In A World Without Frontiers where are all part of the one Brotherhood and The Sisterhood of mankind and we promoted it on on Facebook uh it’s still there it’s

A pretty dead page now because we we haven’t had the energy because it took huge effort putting things up in Arabic and English every day it got um it got up up to around half a million followers and uh mostly in towns like Cairo and it’s interesting though that

People are eager for a philosophy of inclusivity in a world where there is so much much exclusivity of the other um and so that was an experiment okay but the there are other approaches sorry Debb’s accusing me of talking on too much but uh but I think we really

Need to try and look at this whole thing issue of extremism in a different way uh are there more questions Veronica thank you maybe some online as well I have no suggestion I’m working for the next Century Foundation sorry I got a frog on my throat um I just believe that if you

Discuss things the the whole emphasis is is sorry you don’t get together and discuss things I think if we can actually have a situation where people can talk and and discuss things and they s suddenly find it that person is quite like themselves and they don’t

Put up a war because the well you’re Muslim and I’m I’m Christian and and and you ought to be able to discuss things and I just think it’s so sad that people no longer just have opportunity to discuss things and um and always the this this great um umbrella of the

Internet sort of covers everybody and so you you you don’t meet people and and discuss things I mean there’s the thing is you to meet people and talk about things I think really important to be able to do that and and now in this internet world we don’t discuss people

And and find out that person’s quite like ourselves and and put out walls and and I just find it really really sad that we’ve got to this where we don’t actually meet people and talk to them I think that’s much better I’m sorry I have no no other solution that’s what I

Was going to say um are there any other are there any online questions GW okay so can you thank you okay thank you very much my name is Joe I’m from Hong Kong China uh I’m the chief editor of uh artificial intell Global community so uh I love this topic very

Much so uh my question is should we consider the uh historical perspective when when we discuss uh extremism uh 50 years ago uh lgbtq was not a good thing 50 years ago but today it has become a common sense 200 years ago uh uh colon uh colonis was

A social Pro progress but today it’s a bad thing uh so so today we discuss about how to protect uh children and how to educate them but actually I think children themselves is educated themselves by TI Tok uh when we discuss about Tik Tok uh spread uh

Uh bad ideas but it also share lots of knowledge uh to Next Generation uh we could imagine that 100 years ago uh probably BBC TV is a evil thing it’s not good uh as same as we treat uh Tik Tok today so uh so uh I want to know

Could we uh probably after 100 years uh everything narrative uh will be changed and uh today’s uh extremism extrem extremism will become uh another another thing wow okay that it’s quite a question uh I the one thing I would say about the the BBC there there are regulations uh about impartiality I mean

They’re regulate Tik Tok is not regulated there are responsibilities for public broadcasters certainly in the UK and that’s regulated by ofcom so I don’t think you can sort of compare apples and peets that’s the only thing I would just like to point out so um Paul would you like to tackle

I mean wow extremism is it going to become in a 100 years yeah yeah yeah yeah and I think question thank you so stepping back yes of course ideas Norms values do seem to change over time over history you know we convince ourselves that certain things were true and horrible or

True and right a hundred years ago which then turn out to be the opposite or close to so therefore what’s the solution um is a solution to fight those things or is the solution to actually deal with rues of Engagement so let me unpack that just a little bit

When I’m talking to people with different positions and my positions changed over times as well but when I talk to people with different positions I’m not looking to beat them I we set up Rules of Engagement and Rules of Engagement help us navigate in our in in our

Conversation how we’re going to discuss that so we’re going to be respectful yes what does that mean we have to Define that are we going to be loving Yes sounds fluffy but it has ideas have legs these emot ions have legs if I’m angry and I don’t settle that has legs in

Terms of I’m going to act out of that so if we’re discussing an ideological position which we may disagree on or which may well change in 50 years time I think it’s really good to train people how to actually engage with other people in conversation because we need to have

A a humbler position I’m not saying that certain extremist positions we take all the way through forever but if I’m not a philosopher but if you read Hegel when he talks about an um antithesis and then um well when he talks about synthesis bringing these bringing ideas together I

Think it’s a much better way to approach these ideas which may well change we have a humbler position we’re not rolling over we’re trying to pull out what is genuinely useful understanding we change but I think there are some red lines I think there has to be some red

Lines around the Dignity of a human person and a human being the oldfashioned things about love all those things actually have stood the test of time as ideologies and extreme positions have come and gone they seem to be the themes that cut across so I wonder if we

Focus on the universal themes that cut across are shared humanity and build from that place might be a bit of a pie in the sky thought but I can’t think of anything better William well there’s so much to say I think Veronica by the way was

Quite right about the the um issue of talking to one another that is essential and it can be done too often social media is is a a place where you put your own View and then walk away and then you get your followers but there’s not a proper conversation in every instance

And we need to make social media a place for conversations uh so that when people post something you respond rather than just take it down that’s one of the things by the way um I talked about Kata we we promoted that I we’re talking five 10 years ago now and that was on

Facebook now the the big with the young people we should be on Tik Tok or SnapChat or whatever these are the big programs Facebook is it’s not dead in the Middle East but it’s pretty dead here in the west and uh yeah Tik Tock is where it’s at Tik Tock is what young

People look at and so that’s the the the platform that needs looking at and I don’t think our chair is right I don’t think the BBC is regulated in my view the amount of extreme graphic violence on the BBC that promotes extreme violence is not regulated and I don’t

Think the BBC is balanced I don’t think their coverage we’re here on Holocaust Memorial Day today and um we’re asked to put a candle in our Windows to remember the dead from extremism but it’s very hard um at a time when people are dying in Gaza

Um it’s very hard it’s very hard to to adopt these conventional sort of positions um I I think that we Debb’s pointing out the time I’m I’m I I will I will stop but I think your question is very powerful very important and um in a historical context in a historical

Context you’re right uh the old ideologies we’ve moved on it’s not a world in which uh communism and its traditional um Marxist then this sort of idea well that it it died pretty swiftly traditional communism it as an idea iology I’m not sure you you got it under

Stalin anymore um it ended pretty swiftly but uh there are these ideologies are gripping you know the ownership of the means of production by the people for the people Clause four of the labor party we do ideologies grab people whether it’s Isis communism whatever and um and we need to be aware

Of that and you don’t fight an ideology by trying to repress it but the historical context yes it does matter I need to repeat the BBC has very strict guidelines and Regulation and and that’s a whole other discussion but I think on on impartiality we need to um wrap up I’m

Going to we’re all right are we all right oh do we need to wrap up yeah my goodness I was looking at the clock up there yeah so um I’m just going to say that I think as you’ve seen the chair disagrees with some of the speakers um

But we need to continue this conversation we could go on and on I think this is a subject we will revisit and I think the big takeaway for me is that obviously education is key and that governments I think politicians also we haven’t touched upon this their language

Of division and hate also sets a tone it does that’s true and and so I’m sure the speakers will be happy to speak to you afterwards um if anyone has any further questions so I’m going to wrap this uh discussion up and thank you very much for coming today thanks you

Mic check one two for for for for for hello hello hello hello is anyone can can you hear me hello can you hear us hello yes can you hear me could you repeat that sorry yes can can you hear me yeah we can hear you thank you okay thank you

We’ll be going live about five minutes all right thanks all right hello uh recording in progress oh cool um welcome welcome to the afternoon panel on investigating cyber crime against business in the EU results from the cyber security for businesses and we like to welcome our speaker uh Tomaso

Kamale and yeah thank you yeah thank you you may begin now thank you yeah perhaps I share the screen yes please uh yeah you should be able to see the we can see you perfectly okay nice uh well first of all good afternoon everyone and thank you

For for this opportunity to present a study and a project that um I’ve been developing over the past few years together with my colleagues at the center for the of democracy which is a public policy think Bank located in sofhia Bulgaria my name is Tomas kumal and I a security analyst and project

Manager who has been following this project on cyber crime against businesses um in the Netherlands Bulgaria and Spain which I will tell you more in a few minutes um essentially we developed a survey instrument uh so a research tool to detect cyber crime and cyber victim victimization among

Companies in the three Target countries uh because we we found and we we noticed in the literature that despite the increasing presence of businesses in the online sphere in the past decades and and increasing cyber crime and cyber security threats that um have been uh targeting companies and

Individuals only few studies had had developed um a new level investigation of cyber crime against against companies in in you States um So based on our on our studies we we found that there was a need of a CrossCountry cyber security survey for businesses uh that could

Measure the the real impact and extent of cyber crime and the factors that make companies more likely or less likely to report cyber victimization to law enforcement agencies or relevant authorities uh the truth of the matter is that today there is now a study from the European commission which is called

Eurobarometer flash eurobarometer which which has also started to investigate these these type of threats against companies um so back then we we decided to to devote our effort and um and and studies to to the to the development of of a cyber security survey for for businesses uh which would have three main

Objectives the first one was to investigate cyber crime and cyber victimization among companies in Bulgaria the Netherlands and Spain which as I said were the three Target countries of our EU funed project uh the second objective was to analyze the reporting behaviors and the associated cost of cyber victimization uh to

Identify why companies were reporting or not reporting cyber cyber victimization and the last objective was to inquire about the cyber security measures and strategies that companies had developed or implemented in in their business domain so with these three objectives we we developed a a novel uh survey instrument which would have the the

Three targate countries as as already mentioned uh each each country would consist um let’s say each country would have a sample consisting of 400 companies um which we designed as a rep National representative s which me means that through computer assisted telephone interviewing we we targeted all the universe of companies including micro

Small medium and large companies and we use as a reference perers the the standard reference period which is usually used in the surveys which is the previous 12 months so we asked for to through a subcontracting agency uh Within each country we asked 400 companies what was the Cyber

Victimization they had experienced the cyber security measures they had been implementing and the type of cyber threats and cost they had they had incured in the previous 12 months so with this uh let’s say overview in mind um we we deployed the survey we collected the data we analyzed the data

And we we published the the results on on on the project project website which is cyber. and what we found across the three companies the three countries uh was that on average almost one in two companies um of the 1,200 companies that we surveyed was had been a victim and a

Target of fishing attacks in the previous 12 months uh and as you can see from this visual representation fishing attacks in the middle of the graph are the the most relevant let’s say uh type of type of crimes um 177% of companies on average were targeted by malware which is different from

Rare and more slightly more than one in 10 companies was targeted by dos attacks and business identity theft was also quite prevalent in one in 10 companies um so we we Ed the the standard categories of of cyber crime from the previous literature to inquire about this cyber cyber victimization so it was

Uh it was quite an interesting result although on a very descriptive level to notice that um fishing attacks malware were the two most recurring type of threats reported by by companies um when I say companies I I mean that we we targeted um we asked uh through the

Survey either it experts or business rep Representatives that had a knowledge on on the it infrastructure so we we only um surveyed people who were in The in a position of providing effective knowledge and information about the cyber cyber threats and attacks then regarding reporting to the relevant authorities because the one of

The other key objectives was to not only to inquire about cyber type c cyber crime against companies and the type of cyber threats but but was also to inquire what to which type of authority companies were reporting to having suffered from cyber crimes and these

Graphs shows uh on the on the left the type of cyber crimes and separate threats and on the on the horizontal AIS um the the type of relevant authorities that were reached out to report any any cyber cyber offense and as you as you can see uh website administrators and

Internet network service providers were the two most recurring um author relevant authorities that were reached out so essentially the the larger uh the larger the the square the the more the number of companies that declare to have reported to that type of authority some other some other authorities as bank or credit card

Company on the top level were mostly report reported as relevant authorities for attempt of accessing a company’s account which is no surprise and another Trend that is partially visible is that ransomware in the middle of the graph was basically almost evenly reported uh by companies when it comes to website

Administration internet and service providers but also the police and Outsource cyber security providers so one of the takeaway of this of this graph and this type of information is that is interesting to notice that the police which is actually one of the state authorities that would

Be mostly um let’s say involved in this type of U this type of dynamic was actually not so much involved uh compared to other types of relevant authorities um then talking still about reporting Behavior we inquired about the the main factors for um for reporting um and among among the Bulgarian companies

The the three most recorded reasons were to avoid or or prevent repeat cyber victimization so more than one in two companies declar to having reported cyber crime uh because of prevention of of of additional type of offenses uh but also to support the authorities in fighting incidents like

That in the future uh almost one in uh four companies and then the the belief that it was the right thing to do was also reported by one in three companies among the Bulgarian sample um for the Dutch and Spanish companies we we also found that legal obligations and mandatory

Reporting together with preventing repeated victimization were two two of the reasons mostly reported for having reached out to the authorities um but also seeking help with the with the incident to prevent further losses and again the moral let’s say idea that it was the right thing to do um so this

This type of these type of reasons um somehow point out to the fact that prevent prevention of repeated victimization but also the belief that it is the right thing to do was uh was one of the factors for which companies uh reached out to the companies uh

Although you can you you can see from the from the percentages that and also from the previous graphs that it was it was not so much prevalent as one would imagine but still there are uh there are some some good takeaways from from this Initiative for for what concerns uh

Under reporting to the authorities so the reason the most recurring reasons for which companies did not report to relevant authorities across the three countries there were three key factors U the first one is that they believe the companies thought that the police and other authorities could not do anything to

Help and you can see at the top level of the of the table uh so for example for what for what concerns ransomware uh 100% of the companies in the Netherlands thought that authorities could not do anything to help so that was the reason why they did not report

The uh the incident to to the to the authorities um but also and this is also relevant uh the two other key reasons for not reporting where that companies dealt with the incident internally so they they manage within within their own security team the the issue or the

Incident and also because they thought it was too trival or not worth reporting so they they didn’t bother to uh to reach out to the authorities to seek help or to to to make a a crime reporting file this is also interesting because in the literature there was also

The fear of reputational damage as one of the key theoretical reasons for not reporting cyber offenses so the notion that companies would be concerned about uh an impact a negative impact on their reputation if the public opinion became aware about having suffered a a cyber data breach or a cyber security attack

But we found no evidence in our study that fear of reputational damage was a key reason for for not reporting um so lack of trust which is essentially they cannot do anything to help and the reason the fact that they dealt with the incident internally or they thought it

Was too trival to reporting were the two key reasons for for not reaching out uh then moving on to the next slide if I okay so the the conclusions that uh that we that we reached through through the study and through the reports which are openly available on the website um

Is that of course reporting behaviors varies among the three countries and the nether has has come out as the most virtuous uh country in this regard which means that in the Netherlands companies at the highest reporting rates for identity theft money laundering money muing even if companies aren’t being

Targeted directly there’s a good chance that their apps or services are being used to facilitate and coordinate the activity between fraudsters you can go online right now and enlist a hundred random people around the world could walk into a bank and open a a checking account to receive fraudulent payments

And the truth is that even if 99 of the fraudulent transactions were caught before they happened good luck figuring out where the initial order came from and you can almost guarantee that unless you piece together the entire case there’s not a single law enforcement agency in the world that will be willing

To investigate or prosecute it what we have here is a problem that’s growing out of control it’s cross-jurisdictional low risk and highly profitable if it was just a matter of building better detection methods we would not have the problem that we’re seeing today in a 1977 lecture by danella Meadows on the

Philosophy of System Dynamics she said that any time there’s a problem growing exponentially we need to identify the positive feedback loop in the system this is where complexity science enters the picture we need to step back and look at the system as a whole so what exactly is complexity science

It’s the study of the relationships between systems linear systems are additive the output is always the sum of the input 1 + 1 equals 2 complex systems are highly interdependent decentralized self-organizing and sensitive to feedback loops what are feedback loops think of this in terms of a thermostat

You turn the thermostat to a certain temperature the heater or the air conditioner kick on until the room reaches the desired temperature then it shuts off once the temperature exceeds or Falls below the desired temperature threshold the thermostat kick effects on and this Loop continues in the case of

Cyber crime as digitization increases online fraud increases as online fraud increases Financial losses to business increase as Financial losses increase the budget to investigate fraud decreases both of these contribute to decreasing investigation capacity and in the investigation and as the investigation capacity decreases the likelihood of catching froster decreases

Even more emergence occurs when a emergence occurs when a complex system has properties uh or behaviors that the individual parts do not have on their own in the context of fraud detection individual actions or components may not constitute a crime it’s not until their actions are placed together with the

Larger Network that a fraud uh ring emerges machine learning cannot detect patterns that have not yet emerged so here we have a less abstract example of how cyber crime and online fraud take shape in a complex system select any blue box or yellow line in this picture is the element benign or

Malignant if you remove or if you remove it how easy is it for new connections to form so that network uh continues uninterrupted looking at just one component of the system doesn’t tell you very much it’s when these components start interacting that the crime emerges these crimes are unconstrained by

Geographic or dimensional boundaries our laws and policies are 10 to 20 years behind where they need to be in order order to effectively combat this problem just as we can’t fight a forest fire by pouring a bucket of water at our feet combating cyber crime and fraud

Requires that we examine the system as a whole to recap fraud and cyber crime is adaptive frosters adapt their methods to our detection efforts it’s collusive more often than not and this collusion overrides the traditional spectrum of fraud prevention measures such as separation of Duties and oversight fraud displays emergent

Patterns fraud can be visualizes networks these are properties of complex systems not of linear systems now machine learning still has its place in the fraud detection ecosystem but it’s not enough if these models were effective on their own then cyber crime would not be growing exponentially year over-ear broad screens are adaptive machine

Learning models are not classic machine classical machine learning assume that each artifact or transaction is independent of the rest this is a faulty assumption for fraud investigation using the bird flock as an example let us assume that we remove or arrest a dozen birds from the flock which dozen dismantle the

Flock up until this point we’ve been using fraud detection and investigation somewhat interchangeably but they really are two distinct processes the more laborious of the two is the investigation component which requires seil vision of the activity that was detected this is the piece of the process that we need to speed up you’ll

Want to keep the automated detection methods in place but don’t let your teams get too hung up on perfecting detection models because it’s an ongoing effort and the models be outdated as quickly as our utility is realized wherever possible invest more effort into speeding up the investigation

Process and improving the models on the investigation side of the equation we recommend using graph databases for this since they schema list you can incorporate new dat new data points quickly without disrupting your main datab uh database infrastructure okay so now we’re going to take these principles and appli them

In a case study and hopefully provide you with some ideas on how to Stage your own data this case centers on refund fraud it’s a really popular scam nowadays uh they’re countless vendors uh advertising these services online they cost vendors Millions per year sadly these schemes often involve compromised employees counterfeit products compromised

Credentials and ven have hundreds of thousands of customers around the world the preconditions that will cover here on the crime script include data points that these companies would already have in their possession graph models get really complex really quickly but as your teams learn the topologies

To look for that are specific to your domain quering the information becomes much much easier graph networks are ideal for modeling relationships between variables we refer to the objects or entities as nodes and the relationships as edges nodes and edges can both have properties properties can be added to

Nodes and edges properties can also be modeled as nodes think about the Star constellation in most cases you’re going to want to reference the entire constellation so it makes sense to model the constellations one node in other circumstances you may need to reference two specific Stars within the constellation and therefore you

Might model each Star as an individual node and add an edge between them called part of constellation building these models takes experimenting and creativity we recommend that you try sketching your model before importing the data but but for the impatient it’s really easy to collapse and expand properties and

Entities once you get the hang of it particular if you’re well vers at doing database showings and grouping uh groupings in your head so Bob and Rob are nodes are friends or have been friends for is The Edge between them five years is The Edge property Bob likes comedies comedies can

Be a property of Bob’s nodes or comedies can be its own node of a different type you’re going to want to add unique constraints for each node type so that every connection to Bob leads to that single node hopefully we didn’t lose you with that explanation using the facts from

The case I’m going to walk through uh what I perceive as the waste of data connection opportunities and the variables and logic I would use to improve the monitoring reporting and investigation efforts we know that we have the customer the recipient the refund requester and the

Employee they may be one person or four people either way we’ll have a unique ID user ID for each person and since collusion overrides a traditional spectrum of fraud prevention measures such as separation of Duties and oversight shared relationships that cross these permission boundaries will be of particular interest if I were

Modeling this I would start with the two not types employee and user customer employee data is likely stored in a separate database particularly if the service platform is on a different domain or uses a separate application your network systems are like a door each time a user accesses

Your platform the IP address is stored think of every possible touch Point interaction that these users interact with your internal and external systems each IP address should be a unique node uh the first three seg segments of an IP address tell you who the internet service provider is you can

Also Gan the user uh geolocation some data points like timestamps are better left in chambul format you want to look for uh locations where an employee for example shouldn’t be logging into um refunding requesting a refund from the same device where they’re interacting with your customer service agent uh device fingerprinting

You can get the device manufacturer the regional settings from that device do those line up with the GL location of the IP address uh that you’re interacting with the shipping address how often have you shipped to this um specific address does the address belong to that customer does it belong to the

Refund requestor are there any addresses on file belonging to internal users uh matching that location same with IP address device fingerprint um is the person ref uh request who’s requesting a refund somebody who’s logged into your internal systems before uh in the ca this particular case that we’re looking at there were no

Verification of the serial numbers or model identifiers of the items when they were returned uh or even when they were shipped out so logging that type of information uh can help you to identify maybe counterfeit products that are being returned photos of the delivered items uh when deliveries are handed to the

Recipient were not required in some of these cases so maybe implementing a policy and I’m actually seeing this more uh delivery services now where they’re required to take that photo regardless of whether they’re handing it to you or not um match the resolution of the photo with the device that is supposedly

Assigned to the user taking the photo uh um to make sure you’re not getting falsified um photographs uh in this case no secondary audit or tracking was um in place to to check okay who checked these items in uh is there an unusual correlation between the number of items that go missing and

The number of items that are being checked in so this would be the sample model these are the connections that you’ll see don’t be intimidated uh if you’re new to modeling your data is a graph it can be um look really confusing at first but I promise that it gets much easier

You’ll start to identify which of these variables you want to leave as uh in tabular format for your own reporting uh maybe alerts automated learning uh investigation once you query certain topologies uh you can pull everybody from the network who matches that specific pattern and is it becomes

Really simple to filter down through your data so just to reiterate companies already have these data points um but to construct these models you often stop short of analyzing the relationships that could cast the fraud sooner because graph databases are schemes adding or developing new data point this is

Hi Michelle sorry we’ve lost connection to your Zoom are you still there I I can hear you Sharon we’ll wait for Michelle to reconnect but if you’re okay to take over that would be fantastic so the last SC here is data triangulation H and it’s companies um think we lost the

Presentation yes don’t worry uh Michelle’s coming back on okay how do they call it technology it’s all good we’re doing very well for time oh did I lose part here yeah we we had a disconnection for a moment if it’s okay Michelle please could you share your uh presentation again

Please yes give me one moment yeah we lost you right on that data TR that last D data triangulation screen might have been the one before that one right before yeah all so I’ll go ahead and repeat this piece again so just to reate uh companies already have

These data points and uh to construct these models but they often stop short of analyzing the relationships that you catch this type of Fred sooner because graph databases are schema list adding or developing new features is really easy graph neural networks are computationally expensive but Community deection is not Community

Detection um or the the uh algorithms that you would use to identify this type of activity within these graphs once you have them set up uh and again the picture of this graph is a represent representation of the case study walk through as you’re linking your data do relationships exist where they shouldn’t

Are relationships missing where they shouldn’t be how rare is the device manufacturer in the region of that user you can use open source information to pull that uh data in Huawei in the US uh in the US versus maybe uh Apple in India uh do the region settings of the

Device accessing your site match the user stated location in the IP address of their uh location has the device appeared with any other transactions is the address linked to any other customers how often have you shipped to that address uh what’s the return rate going to that address is an employee processing an

Anord volume of transactions for certain product categories um do you have your highrisk product categories labeled to flag unusual activity um emanating from a specific region warehouse employee uh cluster of accounts uh does information from public records match the information that was provided uh your fraud monitoring may be split

Between realign decisions and ongoing efforts uh your investigation teams you definitely want to keep that coordination between them but if you have separate systems uh be sure that you your data is talking to each other so that you can quickly link the activity that’s going on between internal users external users new

Users and on that note I’m going to pass it off to Dr Jones all right very good thank you Michelle for such a comprehensive presentation and thank you Dr Burton for your opening we certainly want to thank everyone for attending this presentation the paradoxical role of data weapon and defense the learning

Outcome objectives were to assist you in understanding the complex and paradoxical nature of data comprehending the landscape of cyber crime and online fraud through the lens of a complexity side and how to enrich data and apply it in fraud detection and investigation our contact information is

Shown on this slide and we’re now happy to take any questions that you may have Michelle if there are any questions can you go back and talk about those um data points I think that it’s very interesting for people to understand who may work in any industries that can be

Connected to fraud and how um the the materials are called out that uh may have high uh interest for frauders and how companies can flag these particular materials uh that are potentially half fraud so we start seeing a lot of Returns on those it gives out an alert

Can you give a little bit more on that yes so if you have um let’s say a high volume say MacBooks those are tend to be a really popular category uh falsified returns if you have um the serial number is always on the box so maybe implementing an automated way to

Track that serial number when it goes out the door when it returns look at where it’s being returned from um the if you’re using UPS FedEx uh I forget some of the other uh ship shipping companies your account should automatically uh enable that you should have an API that allows you to import

The tracking number from those boxes that went out you can import the geolocation the address uh and quickly just match it up with your internal system to make sure that the data lines you can pull in public records the owner information of the property um where something is being shipped how closely

Does that match with the name of the recipient um we have a lot of models in place nowadays that can just quickly parse through thousands of emails and um score how close they are similarity in spelling you know maybe Robert and Bob you’re going to put those groupings

Together um use topic modeling for refr requests does the wording come in the same when they submit um on Amazon for example where you’re given the option to to uh insert a comment with the return reason uh how similar is the wording are you seeing a pattern from a certain cluster of

Accounts where they’re using identical wording and those return reasons um delivery directions so maybe sometimes we’ll say uh use this code or uh leave at the door um don’t leave at the door uh come around back look for those types of patterns and have some automated alerting in place to link okay

This activity this wording these phrases are coming from this cluster of accounts um we’re seeing device fingerprints you know people legitimately send each other you know I I send family members items all the time online that I purchase so look at the time of year you know uh family holiday

Is where families do tend to be together it’s not going to be unusual to see multiple device fingerprints coming in from those location maybe everybody’s uh Black Friday shopping online together and so you’re going to also want to take into account the temporal uh conditions maybe create some variables for that um

Look at the metrics between uh certain employees certain warehouses certain regions have those those stacked up against the other individuals within those regions um within those warehouses um not necessarily related to so if you have a hotline for for reporting fraud um and the same goes for uh tips for the

Tragedies that we see online you know all too often we’ll see headline that oh the the FBI was warned that this was going to happen and so you know it’s easy for us to to sit at home in our smartphone jury bench and um you know lash out that the FBI

Didn’t do something sooner but the reality is sifting through thousands sometimes hundreds of thousands of of tips and identifying which of those is credible it takes time so use topic modeling um you know the the large language model that the development that’s gone into to large language models is so far Advanced

At this point that it’s easy to combine topic modeling with the language coming in uh set up Flags a specificity of details um using entity identification you know are specific names given specific locations create scoring uh for for clusters of information that come in where it might

Be more credible because of the specific detail that’s provided in the complaint or in the tip um and again I know this is a lot to digest for for for uh the audience if you have any any additional uh questions or would like Insight maybe in in how to

Handle your specific uh Cas situation we’re happy to provide that that additional context or um Insight but again these are just some ideas for for ways that rilers can address this issue um how you can link this information within your own organization how to make

Better use of the data pull in as much open- Source information as possible uh make use of those um of public records do reverse lookups in um you know it’s it’s easy to say that that you can never have enough data and in or you have too much data

It’s you’re gonna have to to learn through trial and error what reporting is relevant to your domain um and then be on the lookout for new Modia pandium as you start shutting one fraud ring down another emerges it’s going to be an ongoing effort in um you know folks are

Clever that every every uh almost like it almost seems like every barrier that we put in place gives rise to an entrepreneurial opportunity on the fraudster side uh you know we we require account verification ID verification well you go online and you have countless service services around the world people offering to verify

Accounts for pennies on the dollar so um I’m trying to think of another another example um you have account verification uh they s pre-bundled accounts credentials Ste lugs and the the list of of clever tactics that um they put in place to work around our work arounds uh

Are endless and again I took until we get our policies to a place where this type of in or these types of crimes can be prosecut prosecuted quickly um and we get interjurisdictional cooperation um transferring evidence in the US for example uh in order for the uh evidence to be admissible in court

That the the government where the evidence is collected needs to certify that evidence and that has to go through the office of the um it’s the oia and those those requests can take months or years to process so if and then if we don’t have um uh an extradition

Agreement there’s little incentive to spend the limited resources that that a department may have um you know chasing after something that may never actually result in an arrest or may ultimately end up being transferred to another agency um and at the end of the day your your

Your taxpayers are like hey where did our money go and if you you have to have something to show for it so that’s another uh contributor to the low prosecution rates in these types of cases I would like to bring out another Point uh noted is the device fingerprint srls and

Accidentals uh that is the physical fingerprint but in terms of the device fingerprint can you speak to whether it is called out specifically in types such as the physical fingerprint and how is that connected back to the Frog so the device fingerprint it depends on the the API that you have in

Place Lexus Nexus is the gold standard for a lot of companies um and the device fingerprint that that they capture contains these attributes one of the challenges that we run into with Device fingerprinting now is privacy and anonymity two sides of the same coin so you have Apple and Google both now uh

Implementing anti fingerprinting so where we could have captured the device fingerprint includes the uh user agent the device manufacturer the region Regional settings um the browser the browser version the operating system version these are all details that are captured and the extensions that you have installed the apps that you have

Installed um in some cases it divulges the information WR in the background um the time uh a and the there’s so many attributes it it depends on the service but those are the primary attributes that are captured in a device fingerprint and they are I believe the last uh test like

99% specific to any device unfortunately if you have multiple monitors for example accessing a site from one screen is going to give me a different device fingerprint than if I access it from my same device on a different screen so it one user may have thousands of device fingerprints on file as they

Continue updating their operating system as they’re using different browsers different browser versions and those stack up over time but you learn to you can extract portions of those fingerprints um and Link those together by similarity maybe you want to look for operating system system version we know

How difficult it is to move backwards in an iOS version after an update so if somebody’s on iOS 17.2 and suddenly they jump back to 17.1 maybe you want to create a variable that flags that as an alert um and that would speed up the investigation portion

So you have that flag already in place when you look up these users you look for those um certain Flags or indicators like oh hey this is suspicious um or hey this person is been using Chrome for 88% of their logins and suddenly now they’re using Firefox um those unusual patterns are

Things that you’re going to want to flag in the investigation process okay the anti-fingerprinting with the uh privacy protection methods are making this less useful of a tool so we’re going to have to continue to evolve um identifiers how we identify the the devices and users that are accessing our systems okay thank

You Michelle you mentioned that uh broaders are definitely outpacing us uh in terms of uh their Antics So based on the research that you conducted to what extent did you find perhaps that the fraudsters are primarily using linear or complex models to outpace us so as far as outpacing that’s been an

Ongoing problem um even since before the internet age that is just the the nature of the Beast unfortunately but the collusion piece really came into like um let’s say with the social media is in the availability of uh these massive chat out Discord telegram WhatsApp uh Facebook groups

Where you can basically lock people out let only certain people in and have your activity continue unhindered uncensored um you also have the F line between freedom of speech and you know the ability to to just frequently Post online hey here’s my illegal service anybody who wants to buy

It sign up and you have hundreds of thousands of people in these channels openly speaking about elicit activity um the detection mon they share they know who the different providers are the different detection methods uh that’s another few I forgot to mention that so in implementing your your detection

Methods within your organization you’re going to want to keep what you’re using to detect users or detect this type of activity um limited to only people who need to know because as soon as you disclose it everybody knows the work around um that’s I’m not sure it totally

Addressed the the question but I think it’s just a a an important piece to to mention here yeah thank you thank you it sounded a bit uh more trending towards the linear models yeah so as far as linear models that is you is this is this credit card

Stolen yay n yeah that’s perfect for um machine learning the the classical machine learning because it’s gauging that transaction and it’s putting classifying the risk to that transaction but your risk from a $5 or $10 transaction is minimal compared to identifying 5,000 people who were working together

And you’re you’re gonna miss the the the forest with the trees so to speak okay thank you were there any questions that came in perhaps online or in the room hi we’ve had no questions come in at the moment I’m afraid okay all right so I’ll turn it over to

You Michelle to close us out all right well again we want to thank you for joining us today our contact information here on this slide and I’m going to stop sharing my screen wonderful thank you very much thank you thank you thank you Hey Calvin good morning how are you doing I’m great thank you how are you I’m doing great wonderful uh we’ll just wait a few more minutes um we’re over here in Oxford we’re set up to be ready to go at half three so we’ve still got another 12 minutes left actually

You’re nice and early which is good so um thank if it’s all right with you we’ll just wait until the next uh the other partici ipants turn up we’re waiting for Daryl Sharon Ty and anurada yeah even have I don’t know if they’re going to show up even if they

Don’t show up I I present the presentation wonderful that sounds good to me okay cool we’ll um just go back on mute for now uh as we wait for more people to fill up the room but I will come back in and say hello again in a

Moment okay do you mind if I share my screen now yeah go ahead feel free to uh test it and make sure it all works that’s no problem at all yep so I I’ll stop sharing then and we’ll pick up when we come back on yep

That looks good to me I can see that advocating for human factors excellent cool okay yeah I shall speak to you again in a moment see you in a moment okay you are in the world thank you for joining our session today and as you can

See from the title this session is going to be about we’re going to talk about really bringing human factors engineering to the Forefront and artificial intelligence and so here are some of our agenda items for today so gen we’re going to talk about some human factors engineering and AI

Three things to remember about human factors engineering the goals of human factors Engineers some some of the historical Evolution behind it um designing AI systems the role critical roles of human factors some of the challenges awareness and we’re going to get into some of the key areas and we’re going to talk about

Strategies for advocating for human Centric Ai and then we’re going to go ahead and conclude and then we’ll open up for questions and answers from the floor one thing I love doing about in my presentations is starting all my presentations off with a famous quote and this is a famous quote from a

Stanford professor and she said artificial intelligence is not a substitute the human intelligence it is a tool to amplify human creativity and Ingenuity and which is so true and very applicable to our presentation and so when we talk about human factors we can’t turn on the TV

Today without some type of talk around artificial intelligence and so but the one thing we we’re not really talking about in artificial intelligence are some of the negative attributes or some of the negative consequences of artificial intelligence and on this particular slide what do you see here

You see just some things that was maybe that the media was able to capture like a Detroit man Su police for wrongful detection arresting him based on faal facial recognition technology that fail are we talk about some of the massive failures of artificial intelligence today are the we talk about the top

Eight AI adoption failures and how to avoid them so even though AI has recently went through and grown up on us exponentially there are still some issues with it and this is why the topic that we’re talking about today is so important because while the technology itself is growing exponentially the

Problem with one aspect of the problem with artificial intelligence is that we really having a hard time getting our hands wrapped around how do we make it focus on the user or the human element and this is where human factors engineering and AI comes into play because while artificial intelligence is

Not a novel conell it’s been around since the the 19 the 1950s it’s recently ascendency creates a lot of complexity addressing the human element and so it’s kind of like we all of a sudden we have all these tools all these new toys that’s based around artificial intelligence however we’re struggling to

Really understand what we’re using when we use artificial intelligence and so one of those things that we struggling with the most right now regarding the human interaction and and role with artificial intelligence is where is our place because one of the things we have to look at is how are we going to

Leverage artificial intelligence there are three and not more distinct ways we can leverage artificial intelligence one of it is it’s a tool which it will always be a tool but it is a tool that we’re able to manipulate we’re able to supervise we’re able to manage it to

Help us enhance our decision making to give us some sort of end product or artificial intelligence can be our coworker yes we at the point now where we really have to start thinking about having a digital coworker and the other one is you know in terms of automation where artificial

Intelligence is doing everything for us but we are managing that process so what it’s really doing is displacing the human in the decision making process in that decision- making Loop but it still is a a a area where we have to really focus on how do we integrate AI into our

Space and so and again there could be many more combinations out there I just gave you three so there’s something to think about especially the one when we talking about Artificial Intelligence being a digital coworker so three things we need to remember about human factors engine one it is a

Discipline two it’s a profection and three it is scientific so throughout this presentation I want you to always think about those three things that human factors engineering is a discipline it’s a profession and it’s scientific so some of you might be asking exactly what is human factors engineering and so the international

Ergonomics Association said it’s a scientific discipline with the understanding of interaction between the end users people elements their systems and the and the proffe how it applies to Theory the data the principles to optimize human well-being and Human Performance which in terms help drive better behavior and so this is what

Human factors really is some of you might say well I never seen human factors well think about the last time you you boarded an airplane you saw the principles of human factors at work you saw how the steward on the plane were doing their jobs the dos were in the

Cockpit getting the airplane ready for take off and they were communicating with each other through a through audible sound and so those that’s human factors principles and so the question we have to ask ourselves today is how do we bring artificial intelligence and human factors together and so the other

Thing I want you to think about by leveraging human factors engineering we’re going to enhance the design of our systems our tools our equipment our processes are procedures and which is going to enable us to have more effective and better Human Performance and one thing that um I want you to also

Realize is that human factors engineering is applicable to just about any domain so the question remains this it’s kind of mysterious to me and a question that I I I chase every day as a human factor engineer is how did this discipline get left out of artificial intelligence

And so let’s talk about some goals of human factors engineering since we I gave you the definition so some of the goals of human factors engineering is to help enhance safety and Effectiveness so think about some of the AI tools that we have out there AI capabilities that we have and you think

About are they keeping us safe think about when people go to the airport and the facial recognition technology fails and we have people that’s been falsely arrested or we have artificial intelligence using healthare and it’s giving incorrect information to the doctor so these are some things we

Got to think about is it keeping us safe is it also optimizing human system integration this is so important is where do we fit in into the artificial intelligence capability I want us to think about let’s leverage artificial intelligence as a tool and we working with it than trying

To change the scenario and we end up working for it and then we’ll talk about that in some slides coming up at the end what’s some of the dangers of doing that so human factors also give us the ability to reduce mitigate and limit human errors human errors will always

Exist it’s a part of who we are we make mistakes as people and so that’s not going to change an enhance system availability why does it enhance system availability it’s about the design aspect we are designing systems based on our own limitations and weaknesses and it also increased system reliability

Meaning that because we put the system under so much scrutiny and we use what we call a forcing function in human factors meaning we’re trying to get the system to perform to do certain things even if it’s to do something things wrong in the testing phase we wanted

Those systems to um do things wrong so we can iron out the um iron out the um the adversity of those wrongdoings so we can increase the system reliability when we take it in production and go live with it but one of the things we need to remember about human factors engineering

As a whole is that it is wi acknowledge that it is woefully underutilized especially when we’re talking about investigating issues and developing practical Solutions and this also there’s a list of un unintended consequences around this because we’re not leveraging human factors engineers and this also includes the uh AI space

Um those of you are working at Tech and security space you see this with cyber security same thing we are struggling to understand the human dimension in cyber security and we’re seeing the same thing in in AI now and part of that is because it’s locally underutilized and not

Recognized by security and Technology professionals and so this is why it’s so important for us to advocate for human factors engineering engineering and artificial intelligence is because one one area is the critical area investigation we have to have a better understanding between the user and the

AI system so how do we integrate it how do we leverage it how do we understand it and so this is very important because at the end of the day systems should be built around people that’s the human Centric aspect rather than people built around systems it doesn’t

Work and we must also address the human um artificial intelligence interplay uh interplay that is ensuring Optimum user experience and system performance um I was very fortunate to catch a podcast yesterday and it was the CTO from Doo and one of the things he was talking about is with artificial intelligence

Going forward is that we have to pay attention to the user experience and one way of paying attention to the user experience is um by integrating user experience professionals into the artificial intelligence life cycle development so that they can help us understand how do we capture the user experience from a

Human Centric uh perspective and so that’s very important the other thing he talked about was the user interface and so from a human factor engineering aspect we have user experien professionals been integrated we have user um UI um user in I’m sorry user interface professionals and work

Happening but the field of human factors engineering is a vast field we need a lot more of those professionals integrated into the artificial intelligence life cycle so we can keep the human Centric and build around human limitations and weaknesses and so here’s a laundry list of just

Some human factors concerns with AI um this could have went on for another four or five slides so what I did I just try to pair it down to some of the more critical issues that we see with human factors concerns with AI for instance ethical considerations under ethical

Considerations you could talk about some of the biases that we’re seeing with the data and the algorithms in artificial intelligence for instance like we talked about earlier the gentleman in in Detroit who had the issue with the law because the facial recognition technology failed um user experience even though some companies and

Organizations are paying close attention to user experience some are not it’s not that important they they feel it’s better to produce the technology rather than design and The Human Experience aspect of it the human machine collaboration this is something that’s been done for many many decades but we

Struggle with with it in and and AI um a couple more here just trust and transparency we talk about um we also need to start talking about the psychological impact of AI um our cognitive workload with AI how is that going to impact us cognitively where typically when we talk about our work

Environment we might be more manually involved in our work involment or we might have some type of semi-automation that we’re leveraging but now we’re talking about ringing in artificial intelligence where it’s going to do all the work for us and so it might some people might feel like I’m being

Displaced and you are you’ve been displaced out of that process but you are still a part of the decisionmaking loop and and you got we have to find a way to bring that to the Forefront where we keep people involved in that decisionmaking Loop even though artificial intelligence is doing all the

Work and so the other thing we got to talk about is the historical evolution of artificial intelligence it’s been around for more than six decades and it came about through sciences and theories and mathematical logic and statistics and computational neurobiology and computer science that’s a lot but it’s

It’s it came together over six decades ago and now we at a point now where we’re producing at a at a very low scale right now generative AI with the primary objective of replicating human cognitive abilities in other words with creating systems that think like humans and so this slide here I know

It’s a busy slide I just want to point something out real quick and we’ll move on if you look at the where I’m playing with the mouse here this top portion here is the evolution of AI you see the first digital computers came about between 1940s and 1950 somewhere in

Between there right and so if you look down the bottom portion of the slide it talks about human factors and its development in the United States the the Europeans been using human factors and ergonomic well before we started using it in the United States and in the 1940s

Due the military Aviation but you can see both of these fields were pretty much birthed in the United States around the same time what’s concerning though is that they went on their path of materialization and and maturation the problem is that they haven’t really crossed we’re now just talking about

Crossing the crossing the cross roads between artificial intelligence and human factors uh engineering that’s all I wanted to um point out from that slide is that you can see the growth of both of these fields starting around the same time and so the other thing we need to

Think about is how do we design AI systems with human factors Engineering in mind and one of the ways we can do this is we got to think about it from a cognitive awareness perspective meaning that how much cognitive or mental demand is going to be placed on the human and

It’s just not just the human even though artificial intelligence appears to be doing all the work like I said earlier we got to now start thinking about how we going to be integrated and how we going to be working with the artificial intelligence or it’s going to be a

Digital coworker or it’s going to be giv us um some portions of our our work or is it going to be doing all of our work and then we still going to be responsible for the outcomes and the output of that work even though most of is provided by artificial intelligence

Cognitively we have to think about what’s that mean longterm and what’s going to be the wear and tear on the people that the employees involved with that particular system and then we got to look at how do we design it with enhancing decision making and the decision making piece is

Key because if people can’t understand where they are in the process or how to leverage the artificial intelligence as a tool to complement their work or to complement their production or outcome then it’s going to be problematic for them and they probably not going to trust it and the crucial integration

Here is really again we can’t get over the human AI teaming aspect how do we team it together and there’s a lot of literature that right now that talk about the human AI teing piece but there’s not a literature I’m talk about integrating human factors engineering

Into that capitalist to help us ease out the friction from the teaming of artificial intelligence and people in the workplace and so here are some challenges with it and so again we got to talk about natural language understanding everybody at this point have know been hearing about natural language uh programming right that’s

Where we take it tax and is put giving us an outcome but I can tell you from just playing around with some of the aios out there that the natural language process doesn’t always get it right and it’s very difficult to get it right and it’s one of the changes that we have

Coming down the PIP again the CTO from Deo said yesterday that one of the changes that he G that he’s estimating we’re going to see within the five the next 5 to 10 years is that we now we’re using programming languages to interact with artificial intelligence that where

He believes that we’re going to be using H our regular native languages to communicate with artificial intelligence and It produced the Computing language that we need to produce the uh algorithm for the artificial intelligence I think that is a tremendous step I think there’s a lot of concerns there and I

Think there’s a lot of work that has to be done from a human factor engineering point to make sure that we reduce the friction between um the human element and how do we make that happen I think there’s a lot of work that has to be done and again I’m troubled with that

Because from this point I don’t really see a lot of human factors engineering uh professionals integrated into human factors into artificial intelligence at this point uh the bias and the fairness we cannot get over that um we we’ve seen this before this would continue to build

Issue one of the things that I recommend for the uh AI developers is before you send your work out into production is to make sure you capture enough personas uh personas is something we use in human computer interaction when we building a digital interface we like to build a

Digital interface based on the testing of different personas meaning for instance you might have a Persona group of a black men in my age from 30 to 55 or 60 you might have a female group from 18 to uh 30 you might have a disability Group in there you might have other

Ethnic groups in there and so you capture as much as you can about these groups to eliminate the biases or reduce the bias and to make it all up and increase the fairness of it the privacy concerns we can’t talk about that enough even though um a lot of developers would

Tell you that they sanitize the data that they’re using in some of these algorithms and some of but my concern is that the artificial intelligence itself is able to bring the data together using mathematical computations and statistics and so we say we sanitizing it but then again the artificial intelligence is

Able to bring it back together of the bring together the data that we want sanitized so that’s concerning transparency how does it work and and we’re going to get on that in a couple slides here and user trust user trust is something that’s very important because if people don’t trust the artif

Intelligence if they don’t understand it they won’t use it and trust is something that can be very fleeting meaning today they might be using artificial intelligence tool that they really enjoy then if it be if that tool start becoming more complicated and they don’t understand it two weeks two months from

Now they might discontinue using the tool because they no longer trust to people the other one is ethical excuse me ethical dilemas we can’t talk about this one enough because we have excuse me there’s all sorts of Etha sorry about that there’s all sort of ethical dilemas

That’s going on for instance some people are concerned about sharing like we talked about privacy recently talk about sharing pii some people are concerned about just giving too much information some people are concerned about if they include their information in in an algorithm such as like chat GPT that it

Becomes part of the chat GPT algorithm and so and then there’s also concerns if you generate something from from chat GPT or you generate something from a ai2 you know is it how do you go about the the copyright how do you go about protecting it and so there’s all kind of

Ethical dilemmas and concerns around um those capabilities data quality uh again you know we always have a saying garbage in it could garbage out but how do we sanitize the data and how do we make sure that we get good data so that the algorithm is giving us the most Optimum

Output we talked about user interface design we cannot overemphasize that because if users cannot get past the ability to integrate and leverage the user interface they will not use the tool of capability of product at all they just walk away from and so that that definitely is very negative and

Harmful to the human AI interaction security is another one we could talk about security with AI and and and people all day but I would just say this is that once again if with AI capability especially when it gets in the hands of the adversary can the adversary leverage

AI capabilities to reverse engineer to harm us from a security aspect so those are things that we have to start thinking about and those are things that people and organizations are already thinking about especially those organizations who are under you know just a cyber attacks every day and we do

Have that happening and the adaptation of user preferences how do we get feedback from the users to those AI developers so they can make changes to enhance the uh capability AI to or the application or the system that the user is going to be using that feedback loop

Is key that mechanism to inform the system and here’s the thing with AI just like anything else we use there’s a life cycle and that life cycle has to be open to feedback from the users and if that feedback doesn’t get back to the developers then the user will feel like

They don’t no long trust the system and they won’t use the system and one of the growing phenomenons that we’re seeing with artificial intelligence is AI awareness and so this is all about understanding the AI systems how do we understand how do we categorize it how do we deal with

The different facets of awareness around it and this is a this is very complicated because for instance just think about when you use chat GPT or some other AI tool do you really understand what’s happening behind the scenes is is it maintaining your awareness are you m are you

Situationally aware of what’s going on and if the answer is no then you probably don’t have a full level of trust in the tool even though the tool is able to give us great output in some cases it’s still that factor that I don’t understand what it’s doing and we

Also have to remember that AI artificial intelligence awareness is still in this emphasis stage and we’re going to have to continue to uh develop this in literature so that it it can take root and it’s really going to help us focus on three things transparency remember we talked about

Transparency but we’ll get to that in a few communication how do we have better communication between the the developers and the end users not just the end users but also communication between the AI tool and the persons that using it or the organization that’s using it and calibrated trust remember I just told

You these tools are going to morph over time they’re going to become a lot more sophisticated over time and so again it’s going to require us as end users to calibrate our trust whether we trust the tool because we understand what it’s doing or we don’t trust the tool because

We don’t understand what it’s but when we talk about transparency it’s really getting into understanding the black box of the system and what those black box of the system is that’s the algorithm and the data is using in the algorithm to make a decision do we understand the

Decision- making process and is the system telling us it’s decision making how it arrive how arrive to that final output if we can have transparency to understand that that would help us understand the artificial intelligence and so again how do we go about helping bringing about this transparency uh

Helping the artificial intelligence help us understand how it’s decision making because at the end of the day a lot of people are going to be looking to depend on artificial intelligence to gain the competitive Advantage because it’s there it’s a tool now it’s out there and it’s

Not like we can put it back in the box we can’t it’s out there and so companies are saying how do I leverage AI to be competitive ly advantaged the other thing we got to we we talked about safety accountability and explainability which we’ll get to

Here a few minutes and so this’s this this other phenomenon that’s happening with AI it’s called anthropomorphism and I first learned about this in my academic writing I remember one time I submitted a paper and I had a professor told me that he said you know you are violating anthropomorphism right now and

I had to go look that up and that was years ago and what it is is just basically where you give humanik uh traits and characteristics to nonhuman things and so when we look at anthropomorphism there’s four degrees of AI anthropomorphism and so there’s one called courtesy being polite to the AI

And then one the other one is reinforcement giving praise to the AI and Ro role playing with the AI and the other one is building companionship now this is this is where it can really turn to the dark side this is where cyber psychologist come in and say okay we

Need to understand how people are interacting with digital Technologies on the online environment and so on because now we begin to see people have real life conversation and relationships with artificial intelligence and so this is a concern that we must understand and how do we help end users understand that you

Know giving humanik characteristics to AI is good and it’s bad it’s a double-edged sword it’s a double-edged sword in the fact that the good it can be potentially a digital coworker or it could be a tool that help us have higher Productions and output but the negative

Aspects of it is when people go too far with it and they start to develop companionships and other things with with artificial intelligence and so how do we get that under control and this is where I think human factors engineering can come in and help build some boundaries and help put out some

Communications and understanding that at the end of the day artificial intelligence is still a to and then we you hear me talk about this several times and now we’re going to get into it explainability and AI system we really have to understand the system and for me as an end user to

Understand how and what I’m using I need to understand it beyond the technical aspects of it like again I pretty much understand that chat G PT is using data it’s running an algorithm and then through that algorithm I ask a question I do some prompt engineering and it

Gives him a response I still need more transparency on it because I want to see the decision making process behind it of how it derived at its outcome and so the explainability in AI system is key because if the system cannot explain how it reach its decision or his output then

Again the trust is going to be less people are not going to be trust uh trusting of these systems because they don’t really understand it and so this challenge is this is a real challenge for us as human fighs Engineers because now we have to figure out how do we help

Systems design an explainability aspect and that explainability aspect has to be across all the different personas all the different user groups out there and including my 70 plus year old parents right and so think about that from a developer having to explain a right Cod and right uh part of the algorithm where

It’s explaining this decision making so users can understand it and so that’s going to be a a real difficult challenge but it’s not a challenge that should beond be Beyond us we should be working to do that anyway because now with us moving into more of advanced generative

AI people are going to have more concerns and yes even in Healthcare doctors are also at um excuse me doctors are also asking for explainability in AI systems because now they want to show the patients how artificial intelligence Rees decision and which the doctor is going to most likely leverage in helping

Inform the patients of a diagnosis of something and so insurance safety in AI this is this is very important when we talk about safety you know a lot of us you know think about self-driving cars and this from this perspective think about some of the uh transport carriers on the highway and

That’s going to be uh driven by Ai and now we’ve got to make sure that we got a safety assurance aspect to it and not only do we need just the Assurance we need safety cases where we can ensure that the artificial intelligence is going to keep us safe and it’s going to

Be reliable to do the right things at all times you know I was reading something recently about self-driving cars and most of us already know there’s an issue with self-driving cars especially making left terms and so I find that very fascinating that researchers and scientists are still trying to determine what causes vehicles

To have problems when they uh make left turns now I also was able to talk to a coworker of mine who was able to travel out to San Francisco a couple months ago and participate in the self-driving car experiment and the cars were kind of like imagine a uber like but it wasn’t

Uber where you get in a car there’s no driver and he was able to participate in that experimentation and he was very comfortable with it you know the feedback he got was the car performed amazingly well and the car did a lot better than what I thought it was going

To do because there was no emotion involved in it so the car was just able to do exactly what it was program to C provide a service to the writer and so that’s like why safety is uh so important in this aspect the other thing

We have to admit is that we’re going to make errors the artificial intelligence is going to make errors can you imagine a self-driving car and a on rush hour traffic in Los Angeles or La I mean not La I mean Dallas or Atlanta or Chicago or one of the other major metropolitan

Cities around the world that could be very problematic especially during rush hour so it this also be frustrating but again that is something that we’re going to have to work through and learn to accept that AI is going to make errors and because it make errors we’re going

To have to find a way to reduce the amount of errors it’s making and we going to have to also reduce the amount of errors that we’re making or implementing into the process so that the in case the AI is kind of like a digital coworker or we’re controlling the AI completely

Right and then how do we integrate safety in human FES engineering this is a natural fit one of the goals of of human FS engineering is to enhance safety to focus on safety and so we’ve seen this when you look at some of the other domains that leverage human factors engineering remember human

Factors engineering was born out of mil military Aviation so the aviation industry in the United States and around the world you know very very uh good at implementing human factors in all of Aviation so it really could be done and so we just got to find a way to make it

Applicable to artificial intelligence so that we can in in we get to a high level of safety with um um not so many mistakes and not so many Adverse Events happen then there’s a moral accountability and complex ecosystems so imagine you working for a large bank or you working for a very large

Organization and just imagining your cyber security ecosystem your cyber security ecosystem is massive and so now we’re talking about integrating artificial intelligence into that system and so how do we integrate artificial intelligence into a already large ecosystem and we we’re able to achieve more accountability and that is

Something that we’re going to have to take into account as human factors Engineers because one of the things we look at we like to look at systems as a social technical system especially like cyber security or Finance systems or healthare systems or rail systems or Aviation systems there are not just one

System there are many systems that are dependent on each other and so when you impact one system you could be impacting all the system and so how do we achieve moral accountability when we’re talking about integrating artificial intelligence into a very complex ecosystem and so in some cases we’re

Going to be redefining moral responsibility again because we don’t know to what degree we’re going to be dependent upon artificial intelligence but we just know that artificial intelligence is going to be a part of the work environment so how do we redefine it you know in terms of creating legislations and creating

Regulations and laws to help us redefine more responsibility and then a dynamic Insurance model like how do we ensure that all these things coming together is going to give us the reliability and going to give us the high level of performance that we need how that

However that models work is just yet to be known because we just don’t know again if I had to say where were we on the scale of artificial intelligence specially generative AI till we at some of the advanced stages I would say we probably just now hitting two so we got

A long way to go but we this is the time we need to start thinking about how work environments are going to change due to artificial intelligence and how do we go about creating a d a dynamic Insurance model so that we can achieve the outcomes we want more positively and

Favorable for organizations and not just organizations but for people as well the other thing is human performance degradation um we seen this before and we’ve seen this a lot we just haven’t seen it I talked about it from the AI perspective I remember when I wrote my first dissertation and November

2013 I was the happiest man on Earth because the Federal Aviation Administration put out a report that pertained directly to my dissertation that talked about Pilots leveraging automation commercial polish leveraging automation so a decline in basic airmanship skill meaning flying the air crft and so it fit hand in hand with my

Dissertation and so I was very happy but I say that to say this when we when we leverage automation we have to offset the degradation in our skill sets and so the same thing with artificial intelligence how do we offset the degradation and some of the decision-

Making abilities and some of the skill sets we we used to do in leverage on a routine basis now artificial intelligence is going to be doing it but how do we maintain and keep our skills at at a high level that is something where human factors engineering can come

In and help the company determine it might means we have to create more simulator like environments so that software developers can practice their skill often or depending on doctors who’s going to be leveraging can practice their diagnosis without leveraging AI or whatever however we going to use artificial intelligence so

It depends so we we have to avoid the human performance degradation through leveraging artificial intelligence and so um in concluding here how do we get to a point so we can have human factors engineering and AI there’s some strategies some things I put together believe me this list could have been

25 initiatives or more I just kind of came up with eight here that uh that’s come very common in the literature one of the things we need we need policies directing human CNET and human factures engineering and AI companies are not going to just volunteer invest and a

Hire human factors Engineers unless in some cases they’re directed to do so why because there’s a cost aspect and the other reason is because like we saw on one of the slides earlier human factors engineering is woefully underutilized in some spaces and artificial intelligence is one of them U we need to integrate

Human factors and psychology based professional in the AI development life cycle how do we do that again I just alluded to that we need to hire more people and integrate these people into the life cycle these people are bringing expertise or skill set around the human

Element how do we enhance things for the users and we need to make sure that we’re leveraging uh their capabilities human Center we need to implement human Center and user Center design we say that a lot you and I’m sure everybody has heard that and they’re like what does that mean it

Really means it’s focusing on the requirements of the enduser the people that’s going to be at the center of that you building the capability the AI tools for and how they going to be leveraging looking that the problem from their from their lens and making sure that we account for human weaknesses and

Limitations usability testing we need to test test and retest and the reason we need to do that is because again these AI TOS are going to morph they’re going to advance and so if you stop the usability testing then you you you reduce the chances of having those improvements or those issues raised

Because you no longer doing the usability testing we need to be intentional about and we need to have continuous and be intentional about explainability and transparency that’s a given we also need to develop and teach courses on human factors engineering and Ai and at the Collegiate level there’s

Not enough of these courses being taught and I know some people say well we teach human computer human computer interaction I get it but human computer interaction teaches a very scal down version in human factors it doesn’t teach the full scope of human factors engineering and that’s what we need to

Teach um leverage end users advy groups meaning let end users advocate for the things that they going to be needing in cyber sec I’m sorry artificial intelligence and develop and Implement a human factors Council in AI develop a group of workers and include an Executives and leadership in there so

They have some teeth so that they can advocate for human factors issues and artif artificial intelligence at your organization and and in concluding here just some points IID like to reinforce is that human factors engineering is scientific it’s a discipline and it’s a profession human factors engineer reinforces safety performance and

Behavior AI is a tool that compliments us we got to focus on trust and explainability and we have to ensure moral accountability when things go wrong we got people step up take ownership and mitigate the the wrong so that we can get back to having a system

Operating at a high level very ethically and operationally and with that I’d like to open it up for questions perfect thank you very very much for that presentation it doesn’t look like we have any questions in the room and there are none on the live stream either

Well I have a few minutes in case you want to wait I’ll wait online a few minutes in case somebody want to ask a question perfect thank you it looks like we are all good here no one has any questions um otherwise feel free to close the session early if you would

Like yes we can go ahead and close if there’s no question and thank you for this opportunity no perfect thank you very much it was very interesting presentation and I hope you have a good rest of your day you too thank thank you re sorry to

I’m nice to meet you so um was one of the presenters in um room four yeah and I had to rush out um when you moderated for right yeah do you know the best way to be in touch with her because I don’t think she’s around she must have left right is she

Likely to be at the at the um reception you knows right right that’s the best place to look it up okay so I check that good luck thank you don’t worry that’s nothing you could do that’s just the way it is thank you very much yeah