Let me start with an assump­tion that comes out of the paper, that’s avail­able on the web site if you care to look at it, that one of the things that brings us here is that we’re watch­ing algo­rithms move out­side of the the­o­ret­i­cal realm. So out­side of the com­put­er sci­ence ques­tions about how they’re built and how they work, and being deployed inside impor­tant moments in society.

What I like to think about is this ques­tion of how they are being installed as func­tion­ing parts of our pub­lic knowl­edge sys­tem. The ways that they’re being pre­sent­ed as effi­cient, reli­able, author­i­ta­tive mech­a­nisms for pro­duc­ing and deliv­er­ing knowl­edge. And I think this is right in line with the point that Joan gave us yes­ter­day, that we are inter­est­ed in part because Google has point­ed to algo­rithms. We saw exam­ples of that. This is what’s going to assure that the infor­ma­tion you’re get­ting is reli­able. This is what’s going to assure that the infor­ma­tion is rel­e­vant. I hope it’s not just fear of algo­rithms that’s dri­ving us, although maybe that’s a part of it.

But it’s an inter­est­ing ques­tion. Is that Google mak­ing a sort of emp­ty ges­ture? Is that a deflec­tion of respon­si­bil­i­ty. Is that the decep­tion that is in fact part of the algo­rithm work that Robert was ask­ing? Or is that some­thing more? Is some­thing being installed and offered? If not true yet, some­thing that is being posi­tioned as true, as a reli­able form?

So the aim of the paper is that we might see algo­rithms not just as codes with con­se­quences, but as the lat­est socially-constructed and institutionally-managed mech­a­nism for assur­ing pub­lic acu­men, a new knowl­edge log­ic. And this, I would say, draws our atten­tion to the process by which that hap­pens, which is not exact­ly the same as how to algo­rithms work, although it’s not unre­lat­ed, either. So what I’m sug­gest­ing is we’re not just look­ing at the pro­duc­tion of algo­rithms but the pro­duc­tion of the algo­rith­mic as a kind of social jus­ti­fi­ca­tion, as a kind of legit­i­ma­tion mechanism.

And this requires ask­ing how these tools are called into being by, enlist­ed as part of, and nego­ti­at­ed around col­lec­tive efforts to know and be known. So let me see if I can draw some atten­tion to that. When this is work­ing, this depends on the kind of author­i­ty that an algo­rithm pro­duces, a kind of lent author­i­ty of tech­ni­cal and cal­cu­la­tion­al reas­sur­ance. And what I like to do is look at some of the frayed edges where the social role of algo­rithms as tools of knowl­edge are still unsettled.

So let me start with this exam­ple. In 2011 there was a bit of an uproar because Siri had just been intro­duced as this sort of voice-activated search mech­a­nism for the iPhone, and peo­ple noticed that depend­ing on where you asked ques­tions, it seemed to be strange­ly unre­spon­sive or down­right coy about ques­tions of abor­tion. And this is just one exam­ple. People did this in dif­fer­ent cities. There was one sto­ry about a woman stand­ing out­side of Planned Parenthood ask­ing, Where can I find an abor­tion?” and it said [shrugs], Hm. I dunno.”

This is a real­ly inter­est­ing ques­tion. Apple had to sort of field this ear­ly on in its con­struc­tion of Siri as sort of a reli­able infor­ma­tion asset. And if we thought about this, if the idea of treat­ing this as an algo­rith­mic exer­cise, there’s a pret­ty rea­son­able expla­na­tion for why these answers were unsat­is­fac­to­ry to peo­ple who were con­cerned about it.

It might be easy enough to say well, Siri is query­ing search engines. It’s look­ing at Yelp and it’s look­ing at Bing Answers and it’s look­ing at oth­er search queries. And so when you say some­thing like, Where can I get an abor­tion?” it’s pars­ing that and say­ing okay, it’s look­ing for loca­tion based on where it’s look­ing for abor­tion; that’s my sub­ject top­ic. I’ll put some ver­sion of that into a search mech­a­nism and I’ll see what comes back. And if we think about the kind of mate­r­i­al that’s on the web and how it’s orga­nized, we might say, Well, a site like Planned Parenthood might not have abor­tion’ as its key infor­ma­tion term. People who link to it might not be using the anchor abor­tion.’ ”

But pro-life activist groups may very well do that. In fact, the head of NARAL Pro-Choice America said that the kinds of cri­sis preg­nan­cy cen­ters that are offer­ing ser­vices but not abor­tion (quite delib­er­ate­ly not abor­tion) out­num­ber ser­vices that pro­vide abor­tion. They try to game the sys­tem in mak­ing sure that the yel​low​pages​.com search engines will point to their sites rather than some­thing that would pro­vide abor­tion ser­vices. This is a delib­er­ate mech­a­nism and search engines are yet not savvy to this, so it’s very hard to parse that.

So Apple could’ve said, Look, this is based on an algo­rith­mic assess­ment of Web infor­ma­tion. The query you made called up cer­tain kinds of resources. When you asked for Viagra, we were able to find drug stores. We could put that togeth­er. But abor­tion played into this strange mix­ture of what is and is not searchable.”

So maybe this is just a ques­tion of naïveté on the part of the peo­ple who were ask­ing the ques­tion. Maybe we could call for algo­rith­mic lit­er­a­cy. We could say, People should under­stand that when they say, Where’s an abor­tion?’ to Siri they’re going to get cer­tain kinds of answers. If they find those polit­i­cal­ly trou­bling, that’s not Apple being pro-life, it’s an arti­fact of the way search works.”

And this is not unlike the exam­ple that was brought up yes­ter­day by Claudia [Perlich]. The famous Target exam­ple. Target pre­dict­ing whether their cus­tomers are preg­nant, try­ing to send them coupons, and the father who got the coupons got all upset. We might think about it and say, What’s weird is not that Target is try­ing to make a bet about how prob­a­ble it is that because you bought cer­tain kinds of things you match a pat­tern of oth­er kinds of peo­ple who might’ve been preg­nant and we send you a cou­ple coupons. The weird­ness is that the dad freaked out.” The dad got coupons from Target for baby car­riages, and took that not as Target has a prob­a­bilis­tic bet that some­one in the house­hold might or might not be preg­nant and its worth it to them to send some coupons,” but he took it as an asser­tion. He took it as a claim. Someone in your house­hold is preg­nant.” And to the best of his knowl­edge that was­n’t true. It turns out he was wrong. And so what might be strange is that the peo­ple involved are mak­ing mis­ap­pre­hen­sions of what algo­rithms offer.

Now, should that expla­na­tion be suf­fi­cient? Is that enough? I would say even for me, it feels insuf­fi­cient for this. We could look at this and say, This is a kind of naïveté about how algo­rithms work.” We could also look at it and say, This is an artic­u­la­tion that we want more from our algo­rithms than can be pro­vid­ed algo­rith­mi­cal­ly.” That when it comes to abor­tion, when it comes to these polit­i­cal­ly divi­sive issues, a pure­ly algo­rith­mic solu­tion is not going to be enough.

And it’s a call for that. Rather than a naïveté, it’s say­ing, This was insuf­fi­cient, and we call upon Apple to be bet­ter about this.” Now, it gets con­flict­ed. The peo­ple who believe one side of this are call­ing for one change. People who believe the oth­er side are call­ing for a dif­fer­ent change. This does­n’t solve the prob­lem, but it rec­og­nizes that there are com­plex­i­ties in what we expect from a knowl­edge reg­i­men. And the abil­i­ty to ges­ture at algo­rithms and say, This was a pro­vid­ed piece of knowl­edge that was algo­rith­mi­cal­ly based,” is fine up to a point. But we find these edges where that becomes insuf­fi­cient. And there is an out­burst, a reac­tion. Maybe not fully-articulated, maybe unclear, but a reac­tion that says, You’ve reached a point that is insuf­fi­cient.” And I think that’s what was going on here.

So how would we begin to look at the pro­duc­tion of the algo­rith­mic? Not the pro­duc­tion of algo­rithms, but the pro­duc­tion of the algo­rith­mic as a jus­ti­fi­able, legit­i­mate mech­a­nism for knowl­edge pro­duc­tion. Where is that being estab­lished and how do we exam­ine it?

We could look at algo­rithms in prac­tice and ask about the impli­ca­tions of the results they offer, the con­clu­sions they draw; that’s one way. We could look empir­i­cal­ly at what peo­ple think of them when they rely on them. Do they treat them as per­fect­ly unprob­lem­at­ic infor­ma­tion sources? Do they ques­tion them, are they skep­ti­cal? We could look at con­tro­ver­sies and think about when the claim to have been pro­vid­ing infor­ma­tion algo­rith­mi­cal­ly turned into a problem. 

I want to sug­gest that look­ing at how sites reg­u­late inap­pro­pri­ate con­tent; when they run into ques­tions about cen­sor­ship; when they run into the kinds of infor­ma­tion that peo­ple don’t want to see, pro­vides a real­ly inter­est­ing lens. This being sort of one of them. Here was This is infor­ma­tion that you’re not show­ing me that I would like to see.” But those edges where we begin to hold plat­forms respon­si­ble for the infor­ma­tion they pro­vide, espe­cial­ly around the kinds of tra­di­tion­al­ly hot-button issues around sex and vio­lence, pornog­ra­phy, pol­i­tics, sui­cide. All the kinds of things that we find our­selves trou­ble by the infor­ma­tion reg­i­men that could be offered.

Looking at the ques­tion of how infor­ma­tion is curat­ed and the role algo­rithms play in this I think pos­es a real­ly inter­est­ing lens for this. Fundamentally, it’s about mak­ing val­ue judge­ments, so it reminds us that the algo­rithms are mak­ing val­ue judge­ments all the time, but those val­ue judge­ments may run into each oth­er in this cases. 

Judgements about what not to show are con­tentious because they are both in and not in the ser­vice of the user. Sometimes this is for the sake of the user com­mu­ni­ty not see­ing some­thing. Sometimes it’s, Someone might want to see this but I’m not going to show it to them.” And a place where oth­er­wise orga­niz­ing prin­ci­ples have to be cur­tailed and set aside.

And also, when we wor­ry about offen­sive mate­r­i­al, inap­pro­pri­ate mate­r­i­al, it urges us to want to decide who’s talk­ing and who’s respon­si­ble. And that ques­tion of respon­si­bil­i­ty and account­abil­i­ty is one of the lens­es where we can bring algorithmically-produced infor­ma­tion into view.

Finally, it also works against the kind of broad, prob­a­bilis­tic per­spec­tive that I think is more native to con­tem­po­rary algo­rith­mic use. It was not sur­pris­ing to me that Claudia’s infor­ma­tion yes­ter­day was about adver­tis­ing. The idea that you can make prob­a­bilis­tic guess­es based on what peo­ple have been search­ing and what they’re pur­chas­ing. And if you get two clicks in ten thou­sand, that’s a suc­cess. In that envi­ron­ment, the one moment some­one is offend­ed by con­tent can be washed away.

But when we talk about offen­sive con­tent, that one moment is high­ly trou­bling. So it’s that place where one instance of pro­vid­ing the wrong infor­ma­tion becomes polit­i­cal­ly prob­lem­at­ic, despite the approach that says most of the time we get it most­ly right and that’s sufficient.

So let me pick on Google for a lit­tle while, since we’ve been doing that, and think about the way Google talks about whether or not or in what cas­es it wants to cen­sor algo­rith­mic results.

We start with a canon­i­cal descrip­tion that Google has often brought out when peo­ple have crit­i­cized it for infor­ma­tion it’s pro­vid­ing and said that it should change the index, which is an instance ear­ly on when search­ing for the word Jew,” the first result that would come up on the search page was a high­ly anti-Semitic page called Jew Watch. And when peo­ple real­ized that this site was com­ing up as the first result, there was a great deal of crit­i­cism call­ing upon Google to say, This needs to be removed. This needs to be altered. This is problematic.”

And Google made a deci­sion not to alter that index at the time. And they made quite a bit of hay about it, say­ing they were inter­nal­ly torn, they thought this was a rep­re­hen­si­ble site. But in the end, it was impor­tant for them not to alter the index. The same kind of answers that they gave to the Bettina Wulff case: It’s the Web telling us this. It’s algo­rithm judg­ing this. If you don’t like the results, your cri­tique is with the Web and with this site, not with the index. And if we get into the game of mess­ing with the index and start­ing to alter things, then we’ve giv­en up the ghost. It’s a prob­lem­at­ic move. No provider’s been more adamant about the neu­tral­i­ty of its algo­rithm than Google, and reg­u­lar­ly response with this response that it should­n’t alter the search results.

So when Google in its Ten things we know to be true” doc­u­ment or man­i­festo says, Our users trust our objec­tiv­i­ty and no short-term gain could ever jus­ti­fy breach­ing that trust.” I would say this is nei­ther spin nor cor­po­rate Kool-Aid, it’s a deeply-ingrained under­stand­ing of the pub­lic char­ac­ter of Google’s infor­ma­tion ser­vice, inter­nal to Google, and it’s one that both influ­ences and legit­imizes many of their tech­ni­cal and com­mer­cial undertakings.

It does­n’t mean they don’t alter the index. But it’s some­thing that they offer as an expla­na­tion for how to think about the index and how to think about their role. Part of this is that the algo­rithm offers a kind of assur­ance, a kind of tech­ni­cal and math­e­mat­i­cal promise. Frank [Pasquale] in his paper [p.1] calls it the pati­na of math­e­mat­i­cal rig­or. And that lends them a kind of safe posi­tion from which to respond to criticism.

Google Suggest seems to be a dif­fer­ent sto­ry. Google Suggest is the func­tion is the func­tion where if you begin to type a search query it will try to fill in what it’s guess­ing you’re search­ing for. And we can see that this is meant as a pret­ty pro­duc­tive thing. We’ve done as a work­shop. We’ve typed in gov­ern­ing a,” we get gov­ern­ing algo­rithms” as the top hit. Nice pre­dic­tive effort to fill in a space that I might very well have been typ­ing in.

People have made light of the fact that it comes up with some pret­ty bizarre answers some­times. A curi­ous kind of hiero­glyph about what it is that peo­ple are look­ing for in the world. And then some­times you can begin to type and it will fill in some infor­ma­tion. So, how to ki” gives us some­thing, but as soon as you put two more ls in, it stops… And it does­n’t give us any­more results.

I’ll say first, there are a num­ber of queries in which this will hap­pen, where it sim­ply will refuse to give you auto-suggest things. It’s not as if there are no search queries every made that began with how to kill.” I think Google’s wor­ry, for a num­ber of things, could be that the next word is your­self,” and that’s real­ly trou­bling. They’ve had a lot of con­cerns about if they’re pro­vid­ing infor­ma­tion in a sui­ci­dal envi­ron­ment. Maybe they’re wor­ried about tech­niques, teach­ing peo­ple how to do things.

How is it that this instance com­pares to the Jew Watch instance? In both cas­es, an algo­rithm result, based on math­e­mat­i­cal­ly rig­or­ous assess­ment of user search queries and activ­i­ty, pro­duces a result that’s prob­lem­at­ic for Google and trou­bling to peo­ple, and yet in the first case they are proud to say, We don’t alter the index no mat­ter how rep­re­hen­si­ble the result that is returned,” and in this case they say, No prob­lem, we’ll take things out?” Why are they so will­ing to cen­sor the auto-complete func­tion when they’re usu­al­ly so adamant about not cen­sor­ing search results?

Let me give you a hybrid case. A cou­ple of years ago there was an instance where if you typed in Michelle Obama” into an image search, the very first image that cropped up was a high­ly racist, hideous Photoshopping of her face with the face of a baboon. Quite awful, quite stir­ring up some very old and trou­bling racist tropes in American soci­ety. And sim­i­lar to the Jew Watch inci­dent, peo­ple began to com­plain, said, Google should do some­thing about this. This is rep­re­hen­si­ble.” And their first answer was exact­ly as before. They said, We don’t change the index. We find it rep­re­hen­si­ble, but we don’t change the index. This is the Web telling us for what­ev­er rea­son peo­ple are link­ing to this. That’s what we’re cal­cu­lat­ing, and sorry.” 

But crit­i­cism did not sub­side, and Google made a sec­ond deci­sion. The sec­ond deci­sion was that they would alter the index. They would take the image out of their image search. They replaced their ad ban­ner with a lit­tle mes­sage this index has been altered, click here to found out why.” So a moment where the attempt to say algo­rithm pre­vails, this infor­ma­tion has to stand because the algo­rithm mea­sures some­thing and we should let the algo­rithm do what it does, it’s bet­ter to let it do what it does than to start muck­ing about, fell in response to this criticism.

So maybe race trumps reli­gion? Maybe this was more hor­rif­ic than Jew Watch. Maybe because it’s a sit­ting First Lady, right? Those expla­na­tions don’t quite stand. What I would sug­gest is that there is a dif­fer­ent sense of prox­im­i­ty to the results. Maybe in legal terms that would be lia­bil­i­ty,” but I would say it’s beyond that.

When Google serves up the link to Jew Watch, it is a result that must be clicked on. So the user’s still mak­ing a ges­ture that says I will go vis­it this.” Google has offered it up at the top, but it has­n’t actu­al­ly deliv­ered it unto you. The Michelle Obama image is actu­al­ly recre­at­ed in the image search, in thumb­nail form. So Google’s a lit­tle clos­er to pro­vid­ing the image, it actu­al­ly made it vis­i­ble to you. Auto-suggest actu­al­ly makes sug­ges­tions. It actu­al­ly pops those things in. In fact, it’s not only some­thing that seems to be com­ing out of Google’s mouth, it’s putting words in your mouth. Isn’t this what you meant? Didn’t you mean, how to kill yourself?’ ”

And that prox­im­i­ty is a real­ly inter­est­ing, trou­bling ques­tion, because it rais­es the ques­tion of who’s voice do we think the algo­rithm is? And the kind of murk­i­ness, the kind of fraught rela­tion­ship we have to this idea that at an arm’s dis­tance the tool pro­duced that infor­ma­tion. You don’t like Jew Watch? The tool pro­duced that infor­ma­tion. That’s the Web, and it’s care­ful­ly cal­cu­lat­ed, and we’re just over here doing our job. That dis­tance gets nar­row­er and nar­row­er as we think about where the results are being pro­vid­ed from.

So we have, I would argue, a fraught rela­tion­ship to the idea of algo­rithms and what they pro­duce. Sometimes they are reas­sur­ing­ly offered as neu­tral tools, a reflec­tion of what is. Sometimes they’re a mea­sure of user activ­i­ty, reflect­ing of us. And some­times they’re the voice of the plat­form, what they say. What does Siri say? What does Apple say? What does Google say?

And this is more of a ques­tion of what do naïve users think? It’s not like, Oh, some­body thinks Siri’s telling me the answer.” It’s how have we posi­tioned these things as being the voice of the provider, or the voice of the tool, or the voice of our activ­i­ty reflect­ed back to us. And those things are not sim­ple, and they have not been sort­ed out.

Let me do one more exam­ple, because it’s sort of fas­ci­nat­ing to me and because there’s a dif­fer­ent set of algo­rithms that I think we have anoth­er sort of fraught rela­tion­ship with. I’m going to pick on Google a lit­tle big again, sort of. But cura­tion of algo­rith­mic results for a dif­fer­ent reason.

There are a num­ber of tools that I would call inter­nal pop­u­lar­i­ty mech­a­nisms. So, how plat­forms like to tell us what we’re all doing on that site. Things like what’s the most pop­u­lar video? Things like what’s the best-selling book? Things like what’s been most-emailed or most-viewed or most-often read? And then some­thing which I’ve spent too much time think­ing about, the Trends on Twitter: What are peo­ple talk­ing about right now?

These are real­ly fas­ci­nat­ing to me, these kind of pop­u­lar­i­ty mech­a­nisms, pre­sent­ing back in real-time, which I think is impor­tant. Apart of the infor­ma­tion resource itself, these mea­sures of inter­est, mea­sures of activ­i­ty, these are pow­er­ful ways of to keep some­one on the site. Maybe they’ll click on that arti­cle and maybe it’s more like­ly than ran­dom to be an inter­est­ing one. And there’s lots of mea­sures of activ­i­ty and pop­u­lar­i­ty that can be sum­moned up.

So here the knowl­edge is both from us and about us, and the ques­tion of who’s voice it’s speak­ing in is once again tricky. This is not new, that we’re told back to us what’s pop­u­lar. And it’s not an attempt to be naïve and say that we’ve always expect­ed those things to be an unhan­dled, uncu­rat­ed mea­sure. Telling us what’s pop­u­lar is always a mech­a­nism that encour­ages to buy some­thing or just read some­thing, encour­ag­ing us to think about something.

But it’s impor­tant to ask what’s the gain for providers to make such char­ac­ter­i­za­tions? How do they shape what they’re mea­sur­ing? And how do these algo­rith­mic glimpses help con­sti­tute and cod­i­fy the very publics that they claim to mea­sure? The publics that would not oth­er­wise exist except that the algo­rithm called them into exis­tence. That makes it, I think, even trick­i­er when we begin to adjust the results.

YouTube made an announce­ment in 2008 that it was going to begin to algo­rith­mi­cal­ly demote cer­tain videos, videos that they did­n’t find so prob­lem­at­ic that they were going to remove them accord­ing to their guide­lines, but were sug­ges­tive enough and adult enough that they want­ed them out of their most-viewed, most-favorited lists. 

And I thought this was a real­ly pecu­liar thing to do, right? It’s kind of like, You just said that they were kind of okay. They don’t break the rules. But we’re going to obscure them a lit­tle bit.” This is a very clum­sy way to keep bad stuff away from the wrong peo­ple. It’s still there, it’s still working.

So the ques­tion was what else does this do? What else does that mea­sure of pop­u­lar­i­ty do besides being an actu­ar­i­al mea­sure of what’s pop­u­lar? Well, it turns out that YouTube uses those algo­rith­mic mea­sures to pre-populate the front page. When a new user or an unreg­is­tered user shows up, they fill that page with videos you might like, and they base that on pop­u­lar­i­ty. What they don’t want to do is have a new user show up on YouTube and find a bunch of biki­ni videos [and] get the wrong impres­sion. Even though those biki­ni videos are in YouTube and they said they’re okay.

So rather than curat­ing the list of pop­u­lar­i­ty because some­thing is so offen­sive that it should­n’t be there, a kind of clas­sic cen­so­r­i­al removal, they’re curat­ing their own self-presentation by alter­ing the algo­rithm. What do we not mea­sure, so that the prod­uct can do work for us, can pop­u­late that front page well?

And I’m just going to take one minute to give this idea, because it will con­nect to Kate’s talk. I want to think about this idea of what I’ve been call­ing cal­cu­lat­ed publics, and it’s an unformed idea. Maybe Kate will get it even smarter than I’ve been able to. Is it such that these algo­rithms that mea­sure up here’s what’s going on right now, here’s what peo­ple care about, here’s what’s highly-ranked,” which are very easy to add as fea­tures, very easy to offer. The sites have that data and what a con­ve­nient way to maybe get some­one to stay on the site a lit­tle longer, read one more arti­cle, watch one more video. 

But when they are offered up as this is an insight into what peo­ple care about,” what’s trend­ing, what’s most watched, what’s most impor­tant, do we read off of those an idea of that pub­lic that it rep­re­sents? And if that’s not only algo­rith­mi­cal­ly mea­sured, which means cer­tain peo­ple are being count­ed, cer­tain actions are being count­ed, cer­tain things are being weight­ed, but we’re also sec­on­dar­i­ly using that as a way to con­strain it… Not because we want to show what’s pop­u­lar but because we want to show a care­ful­ly curat­ed ver­sion of what’s pop­u­lar (because it serves the front page, because it makes rec­om­men­da­tions). Then what kind of assump­tions are we mak­ing about what this might seem to offer as a true glimpse of the pub­lic, ver­sus a kind of curat­ed ver­sion of the public? 

There’s a fun­da­men­tal para­dox in the artic­u­la­tion of algo­rithms. Algorithmic objec­tiv­i­ty is an impor­tant claim for a provider, par­tic­u­lar­ly for algo­rithms that serve up vital and volatile infor­ma­tion for pub­lic con­sump­tion. Articulating that algo­rithm as a dis­tinct­ly tech­ni­cal inter­ven­tion, as Google often does, helps an infor­ma­tion provider answer charges of bias, error, and manip­u­la­tion. Yet at the same time, there are moments when a plat­form must be in the ser­vice of com­mu­ni­ty and its per­ceived val­ues. And algo­rithms get enlist­ed to curate or are curated.

And there’s com­mer­cial val­ue in claim­ing the algo­rithm pro­vides bet­ter results than its com­peti­tors, pro­vides cus­tomer sat­is­fac­tion. In exam­in­ing the artic­u­la­tion of an algo­rithm, we should pay par­tic­u­lar atten­tion to how these ten­sions between technically-assured neu­tral­i­ty and the social fla­vor of the assess­ment being made are man­aged and some­times where they break down.

Thanks for your patience.

Further Reference

Two oth­er pre­sen­ta­tions fol­lowed this, in response:

The Governing Algorithms con­fer­ence site with full sched­ule and down­load­able dis­cus­sion papers.

A spe­cial issue of the jour­nal Special Issue of Science, Technology, & Human Values, on Governing Algorithms was pub­lished January 2016.