Skip to main content
Loading Explore guides…
AI, Belief, and Democracy · Acedra
Home
Explore
Philosophy
AI, Belief, and Democracy
Sign in to save
Chat
If YouTube asks you to sign in, open the video on YouTube.
Watch on YouTube
Reload player
Transcript
99 segments
Copy
Hide video
0:03
Welcome, including those joining online. My name is Hazem Zohny, I'm a research fellow at the Oxford Uehiro Centre for Practical Ethics and I'll be chairing this terms Second St Cross Special Ethics seminar which is Co hosted by the Uehiro Centre and the Wellcome Centre for Ethics and Humanities. Before I introduce our speaker, very quick bit of housekeeping, which is to say that this is being recorded for podcasting purposes and it just makes everything easier to stick to a traditional
0:29
format and keep questions to the end otherwise. We need people to sign consent forms and or edit them out later, which complicates things so be easier to just keep questions to the end. OK, onto our speaker. We are very fortunate today to have Professor Mark Coeckelbergh speak to us about AI's impact on our ability to control our beliefs and what that means for democracy. Mark is professor of Philosophy of Media and Technology at the University of
0:56
Vienna and he's visiting the Uehiro Centre till the end of the month.
1:02
Speak for approximately 45 minutes, leaving plenty of time for discussion, after which we have some drinks waiting for us outside of the common area, please.
1:13
Second, to introduce me, yes, but to be here but if you hear a centre where I'm invited as well as here in some gross and yeah, I'm looking forward to your questions and see what you're interested in, in this topic.
1:32
Started off, which is the question is am I bad for democracy?
1:37
And I use various entries to to answering the question and also to clarify. What I'll do today is to you have to focus on this concept of epistemic agency. I talked to this one one way this topic can be approached.
1:58
So yeah, it works good.
2:03
So the, the context of this is that, yeah, concerns about AI and democracy, which also relates to general like, you know, context in which there there is.
2:15
But then see towards authoritarianism, even in countries that pretend to be democracies and in in that context. And there's this new technology and the question is will this is it going to do? Is, is it going to have an impact on that and and if so, how? And then as a philosopher, I want to conceptualise that, I want to to understand it by using concepts and I want to make arguments about it. So that's what I'm doing.
2:42
I started more with with ethics of AI, but at the moment I'm very interested in in using political philosophy to do that. This particular talk connects to to epistemology and and of course moral philosophy. But yeah, that that's a little bit also a kind of context for its talk. So the question is, I bet for democracy of course has to start with with this concept of AI, AI and the various forms of AI.
3:14
Today the the the AI that's very popular usually refers to machine learning, use of machine learning with big data. The computer helps us to find patterns in these data. It's it's basically kind of statistics machine. And yeah in instead of appearing as something that only does calculation, suddenly these technologies, you know engage for example with with language. Use language. One could say famous examples of change GTP and similar.
3:52
But also can be used for for all kind of political purposes, for example for analysing data of social media users and then this can be politically used.
4:07
Democracy is the other concept, and this may you know at least as.
4:14
Vague. If you don't say more about it. And what what I do for the purpose of this paper and talk is to to just make make a distinction between 10 versions versions of democracy which are all about voting. You know, voting when when you have a system representational system with voting, then according to those conceptions of democracy you have a democracy, you're fine. Umm.
4:44
I personally sympathise with take a richer conception of democracy but then what it's it's it's gets tricky but but I find it very interesting. So for example you we have Armas you know this kind of direction things that deliberation discussion is important in a democracy not just voting you need to actively engage citizens.
5:10
So sometimes also comes from from my country, Belgium, and it has an agonistic conception. It's more about struggle and and having your voice heard. These these are all conceptions. I think that that is really not just about voting, but there is a lot about language, also about discussion, which can be rational, less rational and can be emotional also.
5:36
But it's, you know, what they have in common these notions is that the citizens are supposed to engage. They don't only get something, they also have to do something. They have to actively, somehow engage in in their political community by at least discussing this these common questions.
5:57
Another way to talk about this is to talk about political agency. Because if citizens.
6:06
Our love to do that, can do that. Are encouraged to do that. That means that yeah, one can say that as. As, yeah, I'm allowed to to freely vote. That's the kind of political agency. But in this ritual conceptions, political agency means much more. It means that you're, you know, being allowed and being able to reason, to discuss, to defend your view, for example, something that that's normal for philosophers, but it's not not that. Government in political system that we have that that all
6:40
citizens do these things and then yeah being able to to make your voice heard for example is also another way another form of political agency. So it all depends on what you mean by by democracy. What I'm very interested in in for this talk is the conditions for democracy. The the question what is needed as conditions in order to get all these forms of agency and democracy going.
7:07
And yeah, partly they can be about principles such as freedom. If you're, if you don't have some basic freedoms, you're you're in an authoritarian system, perhaps totalitarian system, then, then that's not a you know, the conditions for democracy are not fulfilled, so you cannot even talk about. Let's have a rational discussion, for example, because you're you're in prison or your risk to be imprisoned, for example.
7:36
In Hannah Iron there's also an interesting link to the social what kind of social, you know, environment do we do we need in a society to to have democracy and she makes a connexion with, with with loneliness and and and the opposite of that. So that's interesting. And then what is relevant for this talk is knowledge and therefore also education. And what kind of knowledge do citizens need to exercise those forms of political democratic? Agency.
8:10
And then yeah, we we we need to.
8:16
Look at ways not to talk about that notes and and and one way to do that just one way that there's many possibilities here that one way is to use the concept of epistemic agency which are found in in social epistemology.
8:35
And yeah, I found it an interesting concept.
8:42
Because then it would allow me to to make it more specific, to make more specific. OK, what exactly could be wrong?
8:52
It's political agency then when we look at the conditions for knowledge and and what would be you know could I be bad in some sense and what does it mean? So all my, my. Questions that are asked can now be asked in a more specific way.
9:10
But before I can ask these questions, I need to say something about systemic agency. So in, you know, moral sychology in in in practical philosophy, the ideas that Asians want some control over their beliefs and and have a certain degree, descriptively speaking, a certain degree of control over W's and.
9:36
One can then have discussions, for example, how, how? How voluntary.
9:43
Is that information of beliefs? And in particular in in social epistemology the question is like given that we live in a social environment.
9:54
Yeah. How you know how, how free and voluntary is the formation of beliefs? Because we know, for example, from social psychology, that our beliefs can be heavily influenced by our social environment. So what you know what in that context means this kind of.
10:15
Indeed. Epistemic agency, then.
10:18
What what I do?
10:21
Being philosopher of technology, I I want to know like what about the technological environments or the one could say the technical social environment. So what if we bring in AI there, for example, and see how that impacts the social environment and does that change anything? What is it going to do to to epistemic agency and therefore to to to these conditions for democracy?
10:47
So to make that connexion more precise, one could say that some degree of epistemic agencies needed for political agency in a democracy, especially in the stick version. But even in the in the in the tin version. It tin version means like yeah, I need some control over my degree of control over the formation of my beliefs, because why would voting makes sense if I cannot form my own? Leaves, right? So then if everything is known we could
11:22
just ask AI for example like you know, get. We could collect data about the people have because they are manipulated for example. So this is needed. But also in these versions where where discussion is important for democracy there is very important to. Have yeah. The the the capacity to detach myself from my beliefs reflect on these beliefs and also very important willing to revise them because you know why having a discussion if everyone
11:58
just says like you know this is what I believe and you know that's the end of the of the matter. But the idea is that all this this kind of richer versions of democracy that you also learn something that you are willing. To you know at least for for the the duration of the discussion to to spend holding on to those beliefs at all costs, you know and and we're willing to discuss them. Also the willingness to take responsibility for one's beliefs, willingness to you know
12:36
voice them and and stand for them.
12:41
And then yeah, the the willingness to be openly discussed with these and and revise them, I I like to formulate it in terms of vulnerability, you know, like.
12:53
Due to we render the beliefs vulnerable to reflection and deliberation.
13:00
Umm. So then the question is, with AI, is the development and exercise of these forms of political existing agency and hand, and the exercise of responsibility as citizens possible under conditions shaped by the use of AI and other digital technologies, especially in the social media environments or influenced by social media environments? So this is the question, and my answer is that.
13:28
Since I may influence belief formation in various ways.
13:33
The exercise of forms of epistemic agency needed for political agency in a democracy becomes at least more difficult, and so this constitutes a kind of risk risk for democracy. There's no determinism, because technological determinism would mean that you have a technology and you you have an effect and there's nothing to do about it. This is not how how technology works, and this is also not how politics in the social world works. But it is interesting to conceptualise the influence and
14:07
the risk. So my conclusion will be that AI and dangerous democracy, especially the take conceptions of it and you know, regardless, in other regardless of other ways that I may endanger democracy.
14:24
So let's go into how AI may influence the formation and revision of political beliefs, because that's then what's going on.
14:34
How exactly are these beliefs influenced? And I distinguish between three ways they are influenced. One is a direct and intended 1 the manipulation of beliefs by politicians and the people they they employ. But there's also indirect and non internal effects, one that has to do with knowledge. I'm not so sure about that one. This is something that that I would really like to. Uh pushed honourable to discussion. And then the third one is this phenomenon of of epistemic bubbles, echo chambers
15:14
and so on, which I think also impacts the systemic agency of people and therefore the political agency. So these are the three parts of the of the paper so.
15:28
To in order to discuss that, I picked up a thought experiment from Bondi that's in the social epistemology literature, or at least in the yeah, literature on epistemic agency. So there's a person, Claire, raised in a racist community and comes to believe that people with white skin colour are superior. And then the question is like, how can she, if she wants to change her belief, how can she change? That belief and in in this case towards like a non racist one
16:03
and and the the thought experiment from Bondi is all about social influence. You know this person being in a social environment and therefore having difficulties to change the belief and being raised also in a certain environment. And I'm interested in what about the technological influence? Will Claire be able to exercise her epistemic agency? To change beliefs under the influence of AI.
16:33
So first manipulation of beliefs.
16:40
Social media campaigns based on the analysis of data.
16:45
By means of of AI and data science this is.
16:50
Not, not a thought experiment. It's it's very real. There was this Cambridge Analytica case, happy that it wasn't called the Oxford Analytical case, but I will talk in Cambridge later this spring. So Cambridge Analytica was the case that Facebook Facebook users.
17:12
They they produce data by their behaviour and and these data are then harvested in that case without their consent, talking about consent forms so without consent and also then used for micro targeting people to influence their voting behaviour. So this is very clearly intended use of AI for the influencing the the beliefs of people so.
17:42
So to pick up Claire's example, she's for example targeted by a pro Trump political advertising in order to in that case, you know, not have her change her beliefs, right? So so to stick to her please and vote for for that political direction. So this this case can be interpreted as a case of belief formation revision by means of direct manipulation of beliefs.
18:11
It's somewhat comparable to nudging in the sense that no one is coercing Claire, right? No one is is forcing her to have a certain belief or to not revise her beliefs. But there is an influence under the radars of influence. So it's it's bypassing this, this capacity we have as as restaurant, autonomous beings, something that that philosophers find particularly painful. So this is happening. Means of this technology by means of a I and data science
18:45
and and that one could conceptualise that as saying that class political and epistemic agency is reduced as she is manipulated to change her beliefs or not change her beliefs without her knowledge. So the manipulation she doesn't know about that and without her consent. And this clearly violates the requirements of democratic citizenship the the systemic. Agency conditions for democratic citizenship.
19:16
Then there's the the the second one.
19:21
So here my worry is that.
19:25
Through AI we get a lot of statistical knowledge out there, not always directly but just through the results of AI. For example, if you use chat GP, it's all about statistics of the occurrence of words like that if they occurred together that that's counted and so that's gives a certain statistics and on the basis of that this programme.
19:55
Chris thanks. Now. This kind of.
20:02
Mediation and and yeah bringing to the the public sphere the of the of all the statistical information.
20:12
It's problematic in various ways, I think first of all because it's it's yeah, it's something that's, that's not clear that that's happening. It's it's something that if you look into the phenomenon, if you study it, if you if you're very interested in it, you can find out how AI works.
20:30
But many people use or will use technology without finding out all these things, so they don't even know. You know that this is based on statistical knowledge. But more importantly for here, for this particular worry I have is that people would not be encouraged to look at causal knowledge, causal knowledge, which we get through the sciences. Umm.
20:58
So in this case, for example, there might be statistical information about success in a certain society. It could be that, for example, in American Society that people of a certain skill, skin colour by skin colour are more successful, but that that's statistics then gets true. AI gets amplified and my worry in terms of, yeah, the changing of the leaves is that this. Again, renders epistemic agency, the exercise of that, more difficult because the statistical knowledge is offered
21:37
as as default and I'm worried that people would not be encouraged to look up causal knowledge, causal knowledge which which would tell you that there there is no causal relationship between skin colour and.
21:54
Success in a society so.
21:59
This, this really I think yeah, makes it harder to change players beliefs in the in the in the example. And it's a problem in a democracy if if this statistical knowledge becomes more dominant in relation to the causal knowledge from the sciences. I'm not so sure about this point.
22:25
That it's. Yeah I'm I'm looking forward to to see what you think about the Third Point is also often a way of influencing people without intending so and because although you know some politicians might be interested that people stay in a certain double and just stay having the same beliefs it also happens unintended. So what is the phenomenon? Phenomenon is that if you're engaging through social media and and in social media there is this. Yeah, it's active.
22:58
To to to moderate to to also create this.
23:06
To to analyse the patterns and what happens there is that there's a danger that one always hears the same people, talks to the same people and remains in what's called an epistemic bubble.
23:21
In the literature there's also the distinction epistemic bubbles and echo chambers, and not so interested in that here. For me, it's important here to see this as a problem of epistemic agency again. So if you exclude other voices, discredit other voices, and distrust other sources, then you indeed have this epistemic bubble and you will be less likely, I think, to revise your beliefs and put them. Vulnerable to discussion because you will always hear the same
23:56
for example Claire, there's always this white supremacist echoes and stays in her bubble. And so I think there's a problem not only for the diversity of use as such, but also because people are not encouraged to exercise their capacity for epistemic agency, which is problematic for for this democratic conditions.
24:21
So yeah, this diminishes seasons capacity to render the release filterable.
24:28
To encourage rather people to feel confident, perhaps overcome overconfidence about their beliefs, and this reduces their political agency.
24:40
Again, not necessarily intended. Can be intended, but doesn't have to be.
24:45
And yeah, I contributes to to that.
24:50
By also, you know, moderating social media in such a way that you will see always more posts of the things that you like, you will will see more opinions that you like. And so in the end, you're just, you know, having this this cosy thing with your political friends, but you're insufficiently exposed to other views and even in the ends, don't trust them so much anymore.
25:15
So those are three ways that AI may influence.
25:21
Epistemic agency. Epistemic agency as relevant to democracy.
25:28
And yeah, I think this was one way to to ask the the question of concerning AI and democracy in a more specific way, namely what kind of epidemic agencies needed for democratic political agency. And, you know, the people have sufficient.
25:43
Of that kind of agency, given that AI shapes the agency with of course my normative interest, normative interest to preserve democracy, and normative interest to to then ask about the the relation between AI and and democracy.
26:02
So the answer I've given here can be summarised as follows. Supposing that some degree of and some forms of epistemic agency are needed for political agency in a democracy. In the sense that, at least according to the Tik ideal in a democracy I need to be able to freely form my beliefs, to reflect on my political beliefs, and to openly discuss these beliefs with others and if necessary revise them then the exercise of such forms of epistemic agency.
26:32
These two end up more difficult, if not endangered by AI supporters intended manipulation of political beliefs, unintended defaulting of statistical knowledge as opposed to causal knowledge, and usually unintended creation and maintenance of esteeming bubbles and echo chambers, which discourages rendering these vulnerable to discussion and revision.
26:59
So this is of course a kind of exploratory work, but I think it can be further developed. Can be developed for example, by, yeah, looking all sort of other but related issues that are formulated not in epistemic agency but using another concept. But to also think about like how can democracy be undermines here. And yeah, perhaps also empirical research could for supporters. Claims or or yeah. Refuse some of these claims and for example with epistemic
27:37
bubbles, there's some, also some research that shows that that next to this polarisation and bubbles there's also a diversity of views on social media. And so this kind of research, as far as I see it, is not conclusive, but it's worth and further engaging with them.
27:59
So if this is the case, then you know to to to end with recommendations.
28:07
More towards the outside world is well, if these dangers are identified and analysed are are real, then if we want to strengthen democracy, you know it's it's of course an if you know, there might be regimes and people who don't want that. But if we want to sanction democracy, then we need.
28:29
To regulate we need to regulate AI and and so that these phenomena are less recurring less so there's less risk for these phenomena and we therefore this influences. But also I think education when it comes to nosis is is of course really important. So we need to teach anyway for democracy people how to to exercise their political agency and and create the the knowledge bases for that.
28:58
And that means I think to to you know in this kind of social technological environment when there is more pressure on belief revision and and you know when it's more harder to exercise your political agency. I think in such environments it is extra important to enable and encourage citizens to reflect on their beliefs rather than vulnerable in open and public forums. And and yeah I think this is not just an individual matter to to. And to give individuals the skills, but also yeah requires
29:33
an institutional reform because currently our educational institutions are are not necessarily geared at this. You have to be lucky or privilege to to have that kind of education. So we need more of of that kind of education I think for for everyone if we care about democracy. So in closing I I would like to quote. Uh, one of my favourites, political philosophers John Dewey, who said that democracy is not just a form of government, but primarily amount of associated living. And, and
30:12
there is of course the the one of these take richer ideals of democracy, rights, not just about voting, but it's about community also in in his view. But of course that raises a lot of questions about community. The role of community in modern society and and how how this can be done in practise. But I think for this paper, yeah, this epistemic agency I've been talking about can also take the form of. Yeah, the types of agency that are necessary to to
30:48
make this kind of communicate experience stronger and and see this not only as as a as a means for some abstract kind of democracy. As as as you know, as a valuable thing on itself. Now these these are claims that are stand apart from this paper, but I just want to end with that in order to also disclose the sort of more normative direction where I personally would like to go. However, I think the arguments in this paper are valid for all kinds of political agency and forms of democracy. So for
31:27
technology then I think we need. This.
31:31
Political technologies and call them.
31:35
That enable rather than hinder such forms of living such experiences and even without taking into account that that of course that's that's stimulate and such forms of of epistemic agency have been talking about and rather than hindering rather than than risking to to erode the the the conditions for the formation of such. Yeah such knowledge and such.
32:05
Epistemic agency skills.
32:08
Thank you.
Here to help you learn
Sign in to keep a private conversation with your course tutor.
Ask me anything about the material…
Sign in to chat