Jonathon Penney: I Jon Penney. I’m a legal aca­d­e­m­ic and social sci­en­tist. I teach law in Canada at Dalhousie University. I’m also a research fel­low at University of Toronto’s Citizen Lab and a research affil­i­ate at Princeton’s Center for Information Technology Policy.

Merry Mou: Hi, I’m Merry, and I worked with Nathan over the last year on CivilServant while a mas­ter’s stu­dent at MIT and now I’m a net­work secu­ri­ty engi­neer at a net­work secu­ri­ty startup. 

Penney: So before I begin I just want­ed to first say thanks to Nathan for giv­ing us the chance to come here and talk about our research and being part of this incred­i­ble com­mu­ni­ty that he brought togeth­er. So thank you, Nathan. 

So this is the title of our project and you know, it sounds a bit com­pli­cat­ed. It sounds a bit aca­d­e­m­ic. But actu­al­ly under­ly­ing this project is a pret­ty sim­ple and we think pow­er­ful idea that pro­vides a solu­tion to a com­plex chal­lenge that’s fac­ing online com­mu­ni­ties like Twitter, like Reddit, with­in the CivilServant uni­verse. That chal­lenge is the increas­ing automa­tion of the enforce­ment of legal rules and norms online.

Our solu­tion… No, it’s not send­ing robots to law school. To bor­row Ethan’s line, we’re not that evil and inhu­mane. But it does involve build­ing our own auto­mat­ed process­es, even bots, that can pro­vide a means of pro­tect­ing and reduc­ing some of the neg­a­tive effects asso­ci­at­ed with this automa­tion of legal norms. 

So that’s sort of the idea behind the project and we intend to car­ry it out through the mod­el of cit­i­zen sci­ence asso­ci­at­ed with CivilServant. That is we hope our results and find­ings, which will be gained with hav­ing our bots essen­tial­ly study copy­right bots, and through those insights will pow­er empow­er online com­mu­ni­ties to build their own solu­tions and deal with some of these neg­a­tive effects. 

So that’s the gen­er­al idea. Let me say a lit­tle bit more about some of the specifics of this par­tic­u­lar study. 

So, a lot of peo­ple talk about arti­fi­cial intel­li­gence and the social rev­o­lu­tion it’s going to fos­ter. But when you think about legal rules and norms that rev­o­lu­tion is already hap­pen­ing. Everywhere we look around us, laws and legal norms are increas­ing­ly being enforced and implied through tech­ni­cal, tech­no­log­i­cal, and auto­mat­ed process­es. From very rudi­men­ta­ry police spots, to red light cam­eras and speed­ing cam­eras, to DRM. And in the case of our study, copy­right. Which is the focus of what we’re look­ing at in this study.

Now, as many of you are aware, the Digital Millennium Copyright Act—or at least I hope you’ve heard of it—also known as the DMCA, it’s essen­tial­ly the statute or the reg­u­la­to­ry regime by which copy­right law in the United States is enforced online. It is done so pri­mar­i­ly through DMCA copy­right take­down notices being sent to users, which are effec­tive­ly personally-received legal threats con­cern­ing con­tent that users post online. And it’s a demand to have that con­tent removed. 

What’s hap­pened today, and this was cer­tain­ly not con­tem­plat­ed by the peo­ple who draft­ed the DMCA in 1998, is that now that enforce­ment has become increas­ing­ly auto­mat­ed. Essentially you have pri­vate enti­ties that own and oper­ate bots and auto­mat­ed pro­grams that send out mil­lions, and I mean lit­er­al­ly mil­lions, of these DMCA notices to Internet users all around the world on a dai­ly basis. 

So why is this a chal­lenge? Well, because pri­or research includ­ing research that I have done and oth­er social sci­en­tists have done, have shown that this kind of legal threat that’s received, or knowl­edge that some­one is watch­ing or mon­i­tor­ing you to deliv­er this kind of a notice, has a sig­nif­i­cant chill­ing effect on what peo­ple say and do online. That is, pro­motes a kind of self-censorship. 

Graph source: Chilling Effects: Online Surveillance and Wikipedia Use

This is actu­al­ly a graph from a paper that I pub­lished in 2016 where I used Wikipedia data to show that aware­ness of online sur­veil­lance actu­al­ly has a chill­ing effect on what peo­ple are will­ing to read on Wikipedia online. 

So, that’s a sense of some of the chal­lenges asso­ci­at­ed with this research. The actu­al work, what we’re deal­ing with in this study, this notion of self-censorship is actu­al­ly based on foun­da­tion­al behav­ioral the­o­ries asso­ci­at­ed with con­cerns about sur­veil­lance, con­cerns about social norms. But the point here is that these notices, mil­lions of these threats being sent out on a day-to-day basis, we pre­dict is hav­ing and pro­mot­ing a sig­nif­i­cant cli­mate of self-censorship in online communities. 

Mou: So with that back­ground of quan­ti­ta­tive research and behav­ioral the­o­ries, we want­ed to use CivilServant to answer two ques­tions. The first is, does receiv­ing a DMCA copy­right take­down notice on Twitter cause that user to tweet less often in the future? And we’re hop­ing that this will pro­vide addi­tion­al quan­ti­ta­tive evi­dence for the afore­men­tioned neg­a­tive chill­ing effects that might occur. 

The sec­ond ques­tion we want to answer with CivilServant is, can we design inter­ven­tions, basi­cal­ly on Twitter—Twitter bots—that are based on the behav­ioral the­o­ries that might change how peo­ple react to these take­down requests? 

So right now, if you’re a user on Twitter and you make a tweet, there’s a lot of com­pa­nies and bots who are inter­est­ed in tak­ing down poten­tial vio­la­tions. They’ll send a notice to Twitter and Twitter will often just take down these tweets with­out con­test, due to the safe har­bor pro­vi­sions in DMCA. And at this point the user makes a deci­sion both con­scious­ly and sub­con­scious­ly about how to react to this notice. 

With CivilServant, we want­ed to extend the soft­ware to detect when a user receives a notice by using Lumen, which is a pub­lic data­base of take­down notices across the Web, and you CivilServant to send a tweet back to that user with infor­ma­tion about DMCA take­down notices. This con­tent is designed based off of behav­ioral the­o­ries in the hopes that it will change how that user will respond to that take­down request. 

So this study is still in progress and we’re excit­ed to share the results of the study with you in the future. And we’re excit­ed to be work­ing with Nathan and with CivilServant on this idea that in an ecosys­tem of increas­ing auto­mat­ed legal enforce­ment and per­va­sive Web infra­struc­ture, we as cit­i­zens can still build our own bots that could pro­tect each oth­er from these poten­tial threats to our legal rights. Thank you.