 I'm going to talk about political advertising transparency as a framework of response that has emerged in response to some of the issues that have come up around social media and elections. So political parties have always advertised on different kinds of mass media. And more recently, these days, we also see on social media platforms, there's a lot of propagation of election promises of posturing and so on, and simply understood ads on social media are essentially content that the advertiser pays the platform to circulate more than what it would normally circulate. So in this talk, I mainly focus on Facebook political advertising, which is also quite prevalent in the Indian political context. So what is different or what has changed from earlier forms of advertising. Without making it sound like it is social media is exceptional in that kind of way, but because even when it came to newspapers and so on, there was always this question of how advertising revenue would influence editorial choices and so on. So not to make it super exceptional, but to just understand what is different when it comes to social media ads. And that what you see on the left now, for example, is one of the advertisements, one of the campaign ads of the BJP around the 2019 elections, where it is essentially even in the content of it targeted towards first time voters and talking about how data is cheap under all these government and so on. So typically what we encounter is a video or a photo with a hyperlink that leads outside and maybe a slogan and so on. But what we don't end up encountering is what you see on the right, which is the the political advertising transparency data from Facebook about this ad. And it goes into a lot more detail about who ended up seeing it, what was the demographics, what was the gender, for example, or the regional distribution of the people that saw the ads. So this is kind of what is not very easily accessible for a viewer. Now in 2018, after the Cambridge Analytica revelations, where basically it was revealed that an analytics firm used Facebook products to build water profiles and target users, essentially just do just use Facebook in the way that it was meant to be used. So Facebook had to respond to this kind of public reckoning with their platform. And, and that was the moment in which political advertising transparency was offered as like a solution by Facebook. So in other words, I'm trying to say that it's important to locate political advertising transparency as a response that emerged as a PR move as a damage control and a marketing move by the largest marketing platform in the world. So I'm going to now look at the problematics with political advertising as a category and the transparency as the response. Now, transparency is something that is available only for ads that are classified as political. And the politicalness of ads is determined by Facebook, by looking at two kinds of things one either who is creating the ads or second what the ads are about. Now before of course going into the question of who gets to decide what is political, the very fact that all of these ads circulate in targeted or optimized for engagement kinds of ways, arguably makes all ads on the platform inherently political. But in any case, Facebook essentially carves out this neat category of political ads by creating this kind of binary between commerce and advocacy. So it says that all ads that are commercial in nature are not political ads, and therefore transparency data will not be available about them. Whereas all advocacy related ads will be considered political and therefore there will be some higher degree of information to read about it. And now it sounds basically like a recipe for willful ignorance but in practical terms, this is how it plays out. So on the screen you will see an ad by Shell that talks about more trees, read how nature can play a vital role in the fight against climate change and nature what that means. And basically on the right you have like a community organization talking about encroachment in forests and so on. So the one on the right is of course classified as political and the one on the left is not. There are just for some examples basically a merchandising kind of company that talks about Black Lives Matter would not be considered political. But a Black Lives Matter group that talks about COVID restrictions in schools gets classified and so on. So basically it is in Facebook's interest to keep the numbers of political ads low as the demands for its transparency and therefore intervention into Facebook's advertising infrastructure remains minimal. And in a US Senate hearing I think in 2020 Zuckerberg admitted that less than 1% of the firm revenue is from political ads. Of course in that context he was trying to say that. So I think it was in the context of allowing for misinformation within political ads or something like that to say that they are not doing it in the interest of monetary benefit that actually political ads make up for a very small percentage of their firms revenue. I think what that also reveals is that it is in Facebook's interest that actually Facebook is only classifying 1% of all of its ads as political ads. Okay, so yeah and in India basically there are all kinds and all levels of campaigning that happens. And there are so many actors that straddle these kind of easy binaries between commerce and advocacy that are outfits like news organizations that that on the face of it have a new facade that end up doing different kinds of campaigning that are both level entities and groups and all kinds of things. So, to adapt political ads as a relevant category of governance, even when we're talking about social media. I think is something that I think is a bit questionable. And now I'm going to move on to why transparency also in fact might not might actually distract from what might be needed as a governance response. So since 2019 Facebook has been providing certain tools through which it seeks to provide more transparency. And one of them is the Facebook ad library API. Now this is the we use the Facebook as like ad library API at the ad watch project and basically collect data about political ads. Now there are a number of issues about the accessibility itself of the API and the kind of verification procedures that one needs to undergo in order to access it but I won't get into that. I will just go in a little bit deep into the parameters of data that are available and actually like where those fingers are pointing or what actually that data reveals. So on the left is just like an image of a snapshot of the JSON format data that I captured so the syntax or something is kind of maybe helpful. And on the right is mostly sorry on the right is some of the parameters of data that are available. So the ID, it starts with the I mean I'm starting here with the ID which is basically a unique identifier for each advertisement, the ad snapshot URL leads to the creative that is the visual content of the ad when you click on it. The ad creative body is the text that accompanies a particular ad and link title and link caption have to do with any hyperlink that you put on it. So this is mostly like infrastructural in the sense of creating a different identifier for each ads and things like that. And the second group is essentially the time at which an ad was created and of course the time at which it was created is different from the start and stop time because I can say I can create an ad today and say like only started in February and so on. And then there is the current scene which the ad has been bought the funding entity that has bought the ad, the page unique identifier and name, the publisher platform so like this, and Facebook advertising infrastructure is not just limited to Facebook but also it's family of products that is like Instagram messenger, as well as audience network. Audience network is essentially users Facebook advertising infrastructure to serve ads outside of Facebook products so if you have a, I don't know, an ad in practice or something like that they can for example use audience network for it to serve ads on practice. And finally the language in which the ads are created. So the second kind of paragraph or the second group of parameters are essentially, you could say our advertiser choices, whatever the advertiser has done or is identified with and so on. Now the, what kind of seems to be juicy information of demographic distribution and region distribution is actually a bit of Well, the demographic distribution and the region distribution essentially tells you the outcomes of the targeting process so it's not necessarily what the advertiser intended, but what actually ended up happening. So I could, as Greenpeace create an ad and be like, serve this ad about, I don't know, disappearing mangrove forest to 10 people, and if those 10 people only happen to be young 20 year olds that is a decision that we can see with this data so it's not necessarily revealing anything about the platform choices or even about the advertiser choices but just the outcomes of it. And finally, three parameters in red is in the form of a range so the transparency data is not is not one number but it is arranged that could be like zero 200 rupees has been spent to reach 5000 to 10,000 people and so on. And the relationship between spend and impressions I think is the most is the most important or like the most revelatory kind of thing that Facebook provides but also in the form of a range and not really in a specific way. So all of this to say that in totality if you look at what is actually being revealed. This is not so much how or why I received a particular ad, or how or why an ad was received by a particular person at a particular time chosen from like millions or billions of ads that might be running at that particular moment. It only points towards advertiser choices and not much about the platform, and as more and more advertisers outsource the job of defining audiences to Facebook so earlier what used to happen was, some years back, there was a lot more targeting to say that I don't know target young mothers in the region like Bangalore for example, but now with with Facebook's advertising infrastructure getting more sophisticated, often what a lot of advertisers do is just to say here is X amount of money, I would like to reach why amount of people. How you do it what determinations happen there is none like they don't bother themselves the advertisers don't bother themselves with their information. So this kind of optimization that Facebook does in choosing the most optimal ad for a particular person to see is a decision that still is very much as much of a black box, even in the presence of all of this transparency data, and to that extent, this kind of data might even be like a distraction from the from the actual important stuff. So yeah, compare this with what a Facebook ad with Facebook resident of ads I think said, VP of ads said that there are close to two million data points that go into the determination of why a particular ad reaches a particular place. So that is if that is the degree of data crunching that happens with ML models and so on is transparency the response or is is even is this. That we need to tackle the issue so for example with the model code of conduct that was recently, I think revised, it says, there shall be no appeal to cast or communal feelings for securing votes, but a process of Facebook advertising is by design, and certainly an appeal to different kinds of categorizations whether implied or explicit. And so how does, I don't know, how should, how should, I don't know governance of these kind of advertising platforms we seen when these these are some, like these processes or something that happens by design. So I guess I will stop with the questions of who decides the viewership of particular kinds of political messaging shouldn't be necessarily so individuated and isolated and fragmented. What is the kind of bigger picture that is missed in the presence of only being seen only seeing content that is optimized for engagement and growth of Facebook. What does that mean for rights to political participation and for different kinds of political goals. And I guess the main question of this, the present political advertising transparency framework that is an outcome of a PR move and like a destruction that is something being massively adopted is, is that the way to deal with these issues. Yeah, and this this presentation builds on a body of work developed at ad watch with Michael lead Manuel, who I think is also here. Thanks so much. So while we wait from for questions from the audience. So the one that you've done at ad watch also has looked at kind of international trends in terms of advertising. Could you say a little more about kind of how how they compare with the Indian scenario and what is maybe specific or unique. If there are any such features for the Indian context, as compared to the international context. I think the first thing that comes to mind is that I think three, four years ago, while it was quite newly being investigated like political ads and advertising in general. The US band targeting on the basis of race, gender and a couple of other things for advertisements relating to housing, healthcare. Social security, I don't know like two, three, four things that was only implemented in the US, for example, and at that point, these were not even necessarily categorized as political ads, but now you can see in the, I think in the publications kind of table that Facebook provides that they had to create a separate category called housing and healthcare ads or something like that in order to operationalize this, this principle of not discriminating on those characteristics. So I think that of course remained right just applied there, but what I mean to say is that one doesn't need to be limited to the definitions of the categories of the frameworks that are forwarded by the platforms. In fact, kind of go beyond that, I think. But yeah, it's the first one that comes to mind but mostly also I think even across the world, political advertising transparency has been, I think, as far as, as far as I don't know, regulatory intervention has gone. And within that sense but I might be mistaken about more recent developments. I'm just wondering, I mean obviously we have a very diverse media environment in India, and you know obviously kind of varying levels of literacy, etc. So, kind of, there is a kind of, you know, say a context collapse or a kind of convergence with the way that these ads kind of travel inside and outside Facebook, right. And I mentioned that obviously Facebook's own kind of family of, you know, say Instagram, for example, will be places that these advertisements are shared, but also kind of the relationship with things like, you know, kind of TV news, what's up messages, etc. A sense that, you know, these advertisements actually travel outside the kind of Facebook environment or is it kind of a footprint or impact that's possible to maybe not measure but kind of understand as another kind of vector of influence maybe. I didn't say the Facebook ads necessarily travel out as much as they're part of like the larger fabric of propaganda and misinformation so we know, like for example that a lot of Twitter trends make it to broadcast TV kind of news. And we have, of course, in the more recent reports seen how these are manipulated and so on. So I guess, yeah, also thinking about what Anand was speaking in the previous talk about the focus group discussions to look at what are the kind of optimal targeting narratives to have and so on. So I'm sure all of that goes into creation of these, well, of ad campaigns and so on. So I think I definitely see it as part of the larger media landscape that functions but not necessarily that like Facebook ads on its own have the power to like sway X or Y. And I think from jumping out from that there's a question in the chat, which is, how do you see political advertisement systems go beyond Facebook into domestic Indian platforms such as share chat and cool. That's a great question. I think I think there's a platform design only so different with things like share chat and like you just have like the intro to it is the form is not necessarily, I think advertising so much as much as what platforms like to call organic reach, which is also the algorithmic and not in the trees or something. But yeah, I think, I think just like system design wise platform design. I don't know much about cool, but actually, there was not so much. I will also look at share chat briefly during the 2019 elections. But it was, it was like a totally different kind of landscape, and I think that the ways in which it's used is also much more different was Facebook has for a while been working on its advertising infrastructures that is kind of sophisticated but with share chat there's different kinds of languages and kind of user generated content that then traveled a lot depending on people that picked it up so there's of course also that kind of crowd source in a way narratives that come up. So I don't know how to compare the two, and I don't think political ads in that sense exist on different platforms I think cool. Who is itself an initiative to an extent that comes with comes with its own kind of leanings and that can be said of Facebook to what I just mean to say that I think I find it a bit difficult to apply this kind of political ads framework in its crystallized way across platforms. Right, yes, that makes sense. So I was also wondering, is there a kind of an implication of misinformation slash disinformation, and political advertising. Is there a way in which I mean that's literally otherwise I mean we really don't have any safeguards but can, can one be flagged up as misinformation or disinformation, where is that kind of diagram. I think some years back that was the big debate about whether like misinformation or disinformation that is part of political speech should be censored by platforms and I think Facebook took the one extreme of saying we will allow it and I think I don't recall exactly who it was but I think there was very much a very prominent case in India also many places of political speech containing lots of misinformation. But yeah I think I think they're also the responses have been very much about content moderation centric which is again a distraction from more. I guess algorithmic kind of models and what kind of responses they need so it again becomes this question of takedowns and then about like how do you spot these kinds of speech and it becomes a very it's the it's a computer problem kind of a it's a computer solution kind of a problem. Yeah, I guess that's all I have to say about that. Right. And, oh, and we have a question from here. Do you have a sense of the scale at which political advertising in India is managing to bypass fbs categorization of being marked as a political ad. And what would it have like that looks like so are these for example, more likely to contain questionable content misinformation or disinformation I think this goes back to what you were saying. I think we should stop that something might not explicitly present itself as a political ad but it's it is politically what it is trying to achieve. Great question. I think the problem. Yeah, definitely do not have a sense of the scale but have a sense that it exists but it also depends what one would ourselves consider political ads. There is a lot of advertising that falls outside of the categorization, but we don't. We are not able to measure that because that is not part of transparency data so we're only able to anecdotally look at that within the graphical ad library interface which does not allow for more like systematic kind of understanding. But, but yeah, for example, I think it's quite common for, for, for example, like even coup is a big advertiser I think on Facebook very often and and yeah it's just like lots of funny things happen and my colleague had discovered this that more the heaters that keeps coming into political ads right like the ad about this very old kind of. And but like so much else kind of falls off the map, but it's not it's really, I don't know I think there needs to be some kind of subversive method to understand the scale of things, which was, which I think was possible because understanding what is the number of political ads in total in Facebook, because of the statement by Zuckerberg but otherwise I think I don't know what the scale is, but it definitely is pretty massive. Right, I think we also have an observation regarding and what's apps or relationship to Facebook and what's ups entry into payments via API, let's them have access to real names which are used in bank accounts which are also the same from other banks. So matters keep matters in Facebook's kind of a parent company matters capability after what I think increases. So I think there's obviously a kind of relationship there and I wonder, you know whether that also feeds into the data of, you know, Facebook's own kind of targeting as well. In the API where you can look for different publisher forms, there is. There is also WhatsApp but of course it doesn't throw up any results but at least infrastructurally it is possible or it is kind of envisioned or I don't know how we want to interpret it, but WhatsApp is also kind of envisioned as a publisher platform for ads at some point or like so yeah I'm sure there are other things there happening with data sharing but that's the concrete thing we know that it's part of some backing infrastructure. Thank you so much. We have another question which is, if Facebook is categorizing a lesser number of ads as political, does it run the risk of running into some kind of accountability triangle running with the law or the new social media intermediary guidelines. As far as I know with the intermediary guidelines, I don't think it necessitates political advertising transparency but I'm also not that conversant with it so I'm not sure. But I think Facebook is, I think it's definitely in its benefit to classify fewer as political ads both because it wouldn't have to be transparent about the rest of it as well as any problems that might come within the data that is then revealed about these ads. And I think this will be our last question. Would political ad making qualify as gig work which would make the relationship between Facebook and other platforms and the government requiring reinterpretation. I don't know. I mean I don't think so because political ads, typically like a lot of these content marketing firms do it and I don't think it happens as gig work because it does require. I mean it's not itinerant and I don't think like it works in the same way as gig work. That's just my sense though.