Oumou Ly: Welcome to The Breakdown by the Berkman Klein Center. My name is Oumou. I’m a staff fel­low on the Center’s Assembly: Disinformation Program. Today on the top­ic of doc­tored media and whether they war­rant take­down as a gen­er­al rule, I’m joined by eve­lyn douek. eve­lyn is a lec­tur­er on law at the Harvard Law School, an affil­i­ate at the Berkman Klein Center, and her research focus­es on online speech gov­er­nance and the var­i­ous pri­vate, nation­al, and glob­al pro­pos­als for reg­u­lat­ing con­tent mod­er­a­tion. Thank you for being with us today, evelyn. 

eve­lyn douek: Oh, it’s a plea­sure. Thank you for hav­ing me. 

Ly: Yeah. So an ini­tial ques­tion that was the impe­tus behind this top­ic was, when it comes to doc­tored videos, images, and oth­er manip­u­lat­ed media, what is so sticky about the ques­tion of take­downs, and par­tic­u­lar­ly when the media and ques­tion is polit­i­cal in nature? I think this rais­es ques­tions about…you know, the gen­er­al oblig­a­tions of tech com­pa­nies to soci­ety and the lev­el of respon­si­bil­i­ty we should expect com­pa­nies to assume for the real-world impact of false con­tent that remains on their sites. 

So my first ques­tion for you, eve­lyn, is can you describe the impact of manip­u­lat­ed media both on the online infor­ma­tion envi­ron­ment and in the real world? Just gen­er­al­ly what are your thoughts on the harm that this kind of con­tent stands to pose, and just as exam­ples I mean the sorts of things like a slurred speech video of Nancy Pelosi that cir­cu­lat­ed heav­i­ly last year. There was also a more recent inci­dent from the Bloomberg cam­paign dur­ing the demo­c­ra­t­ic pri­ma­ry where there was a doc­tored video of crick­ets play­ing after Bloomberg posed a ques­tion to all of his fel­low Democrats on the debate stage. And then of course the video of the Speaker of the House tear­ing up the State of the Union speech. So if you pro­vide any insights into that, real­ly would appre­ci­ate it. 

douek: There’s real­ly two cat­e­gories of harm, I think, two buck­ets of harm when we’re talk­ing about manip­u­lat­ed media. And I don’t want to lose sight of the first cat­e­go­ry, which is sort of the per­son­al harms that can be cre­at­ed to like pri­va­cy or dig­ni­tary inter­ests, through the coop­ta­tion of some­one’s per­son­al image or voice. And that’s some­thing that Danielle Citron has writ­ten real­ly pow­er­ful­ly about, you know. Upwards of 95% of deep­fakes and manip­u­lat­ed media are still like, porn. And so I don’t want to lose sight of that kind of harm. 

But obvi­ous­ly what we’re talk­ing about today is more sort of the soci­etal impacts. And that’s still a real­ly real­ly impor­tant thing, you know. Could a fake video of a can­di­date swing an elec­tion? Could a doc­tored video of for­eign offi­cials or mil­i­tary cre­ate nation­al secu­ri­ty risks? You know, these are real­ly live issues and I think it’s some­thing that we def­i­nite­ly need to be think­ing about. 

But the ques­tion also does come up, you know, is there any­thing real­ly new here, with these new tech­nolo­gies? Disinformation is as old as infor­ma­tion. Manipulated media is as old as media. Is there some­thing par­tic­u­lar­ly harm­ful about this new infor­ma­tion envi­ron­ment and these new tech­nolo­gies, these hyper­re­al­is­tic false depic­tions, that we need to be espe­cial­ly wor­ried about? 

There’s some sug­ges­tion that there is, that we’re par­tic­u­lar­ly pre­dis­posed to believe audio or video. And that it might be hard­er to dis­prove some­thing fake that’s been cre­at­ed from whole cloth rather than some­thing that’s been just manip­u­lat­ed. You know, it’s hard to prove a neg­a­tive, that some­thing did­n’t hap­pen when like, you don’t have any­thing to com­pare it to. But you know, on the oth­er hand this kind of thing, this con­cern has been the same with every new tech­nol­o­gy, you know, that there’s some­thing par­tic­u­lar­ly per­ni­cious about it, from tele­vi­sion to radio to com­put­er games. So, I think the jury’s still out on that one. But those are the kinds of things that we need to be think­ing about, and the poten­tial soci­etal harms that can come from this kind of manip­u­lat­ed media.

Ly: More than that, what respon­si­bil­i­ty do plat­forms have to mit­i­gate the real-world harm and not just the harm to the online infor­ma­tion environment?

douek: That’s real­ly sort of the big ques­tion at the moment and sort of the soci­etal con­ver­sa­tion that we’re hav­ing. It’s nice and sim­ple. I’m glad that I can give you a sound­bite answer that will get me into trou­ble with one par­tic­u­lar camp [indis­tinct] I’m sure. 

I think that obvi­ous­ly, we’re in a place where I think there’s sort of a devel­op­ing con­sen­sus that plat­forms need to take more respon­si­bil­i­ty for the way they design their prod­ucts, and the effects that that has on soci­ety. Now, that’s an easy state­ment to make. What does that look like, and that’s where I think it gets more dif­fi­cult. I think we do need to be care­ful in this moment of sort of techlash—which I believe is still ongo­ing, some peo­ple have called it off dur­ing the pan­dem­ic but I think it’s still going—that we don’t over­re­act to sort of the per­cep­tion of harm and cre­ate a cure that’s worse than the dis­ease as well, because there are impor­tant speech inter­ests here. So, I’m not a free speech abso­lutist by any means. I am very much up for liv­ing in that messy world where we acknowl­edge that speech can cause harm and we need to sort of engage in that project. But I do think we also need to not lose sight of the free speech inter­ests that are at play and the good that can come from social media plat­forms as well.

Ly: Definitely. So what you just said kind of reminds me of some­thing that has emerged over the last cou­ple of years, cer­tain­ly since the 2016 elec­tion. And it’s the idea that a plat­form can be an arbiter of truth,” and I think it was Mark Zuckerberg him­self who coined that term. And I thought at the root of it is this idea that mak­ing a deci­sion on whether or not a piece of con­tent, whether it’s false or not should stay on stay on the site, in a way makes that deci­sion­mak­er the decider of what’s true or not. And I won­der… Well first, how would you respond to that, just that notion? And what do you think about that as a jus­ti­fi­ca­tion for allow­ing false con­tent to remain online in some cases? 

douek: Yeah. So I mean, I do have some sym­pa­thy with the idea that these plat­forms should­n’t be and don’t want to be arbiters of truth. It’s not a fun job. And…you know, it’s a good line and I think that’s why they trot it out so often. Like of course we don’t want Mark Zuckerberg or Jack Dorsey being the arbiters of truth. I mean…, …, …come on, right? Like, their truth is not my truth, so…we could start from that propo­si­tion. And you know, it would be like…a ter­ri­ble job, you’re only ever going to upset a whole bunch of people. 

But, that’s not the end of the con­ver­sa­tion. It’s still… I mean it’s obvi­ous­ly an over­sim­pli­fi­ca­tion and a dis­trac­tion from a lot of the issues that remain at play. So, we don’t need them to be arbiters truth, but plat­forms are not and have nev­er been neu­tral con­duits. And they are mak­ing deci­sions about what con­tent to leave up, take down, pri­or­i­tize, ampli­fy, all the time. And so to pre­tend that they’re just sort of sit­ting there, hands-off, not being arbiters of truth is a mas­sive sort of…complete over­sim­pli­fi­ca­tion of the issue. And it’s real­ly only only the begin­ning of the con­ver­sa­tion to say, Okay, you don’t need to be arbiters truth, but you do need to do some­thing. You can’t just be com­plete­ly hands-off.” 

And so like I said, they need to acknowl­edge the impact of their design choic­es. And they need to be much more trans­par­ent about when and how they stack the deck in favor of cer­tain kinds of con­tent, and how they manip­u­late or dis­tort the infor­ma­tion ecosys­tem, which is def­i­nite­ly hap­pen­ing. And we are get­ting to know more and more about that, but there’s still nowhere near enough infor­ma­tion about exact­ly how those ecosys­tems work. 

Ly: So you men­tioned there are a range of oth­er tools that plat­forms have at their dis­pos­al aside from leav­ing up or tak­ing down. Would you mind just describ­ing what that slate of actions might be, might look like?

douek: Yeah, so I real­ly think we need to get out of this leave up/take down par­a­digm, because plat­forms have so many more tools avail­able at their dis­pos­al. They can do things like label things as hav­ing been fact-checked or manip­u­lat­ed, in the con­text of manip­u­lat­ed media. They can reduce the amount of cir­cu­la­tion that a piece of con­tent’s get­ting, or how easy it is to share it, or sort of down­rank it in the news feed or the algo­rith­mic feed. 

They can also make sort of archi­tec­tur­al and struc­tur­al design choic­es that can have huge impacts on the infor­ma­tion ecosys­tem. So an exam­ple here is WhatsApp, in the con­text of the pan­dem­ic, has reduced how easy it is to for­ward mes­sages. So instead of being able to for­ward it to mul­ti­ple peo­ple at a time you can only for­ward it once. And this has reduced the cir­cu­la­tion of cer­tain kinds of con­tent by 70%. Which is like, an absolute­ly huge impact. And that does­n’t involve being the arbiter of truth of the con­tent in ques­tion, but it does dras­ti­cal­ly change the infor­ma­tion envi­ron­ment. So those are the kinds of ini­tia­tives and tools that plat­forms have that I think we need to be talk­ing about a lot more.

Ly: Do you think there’s any use in plat­forms devel­op­ing a uni­fied pro­to­col on take­downs at all?

douek: So I think this is one of the most fas­ci­nat­ing ques­tions. I love this ques­tion. And I’m obsessed with it and I don’t know the answer to it. 

Ly: Okay.

douek: So, when do we want uni­form stan­dards online? And you know, when do we want like, dif­fer­ent mar­ket­places of ideas, so to speak? So I think you can see argu­ments for either. On the one hand if you want stan­dards, you want stan­dards, and you want them uni­form­ly across the infor­ma­tion ecosys­tem. So, devel­op­ing the tools to detect and iden­ti­fy manip­u­lat­ed media is poten­tial­ly extreme­ly expen­sive and might be some­thing that only the largest plat­forms have the resources to be able to do. And if they do do that, why should­n’t small plat­forms also have the ben­e­fits of that tech­nol­o­gy and use the same tools?

But on the oth­er hand, free speech schol­ars get ner­vous when you start talk­ing about sort of com­pelled uni­for­mi­ty in speech stan­dards, and maybe if we don’t know where to draw the line, why not have lots of peo­ple try draw­ing it in dif­fer­ent places and see what works out best? This is some­thing that I’ve been call­ing the lab­o­ra­to­ries of online gov­er­nance” approach to this problem. 

So, ulti­mate­ly I actu­al­ly hope that we can find a mid­dle ground. Like a good lawyer I’m, you know, some­where in between where we can sort of have the resources and some sort of com­mon sort of stan­dards but some flex­i­bil­i­ty for plat­forms to adapt those to their unique sort of affor­dances and their unique environment.

Ly: Thank you so much for join­ing us today Evelyn. I real­ly enjoyed our conversation.

douek: Thanks very much for hav­ing me.

Ly: Thanks.

Further Reference

Medium post for this episode, with intro­duc­tion and edit­ed text