[This is a response to Lucas Introna’s pre­sen­ta­tion of his draft paper Algorithms, Performativity and Governability.”]

Matthew Jones: So I want to begin with a great failed data min­ing project from around 2001. It was was a great data min­ing project that would apply automa­tion to team process­es so more infor­ma­tion can be exploit­ed, more hypothe­ses cre­at­ed and exam­ined, more mod­els built and pop­u­lat­ed with evi­dence, and over­all more crises dealt with automatically.” 

And this project—and I think it’s very inter­est­ing for out point—it rest­ed fun­da­men­tal­ly on a cri­tique of human rea­son, of human indi­vid­ual rea­son. It said that what it wished to fund were A: cog­ni­tive tools that allow human and machines to think togeth­er in real-time about com­pli­cat­ed prob­lems; B: means to over­come the bias­es and lim­i­ta­tions of the human cog­ni­tive sys­tem; C: cog­ni­tive ampli­fiers to help teams rapid­ly and ful­ly com­pre­hend com­pli­cat­ed and uncer­tain situations. 

This was a mas­sive data min­ing effort pro­duced in the wake of what were wide­ly seen as fail­ures with­in arti­fi­cial intel­li­gence and machine learn­ing to pro­duce evi­dence that was­n’t ground­ed on such an under­stand­ing of the lim­i­ta­tions of human abil­i­ty. It went under the name Total Information Awareness. Some of you may remem­ber this. This was a mas­sive coun­tert­er­ror­ism effort. 

And what inter­ests me about this and the rea­son I men­tion it is that it seems that those of us dis­cussing the inscrutabil­i­ty of algo­rithms, as Lucas and oth­ers today have done such a won­der­ful job discussing—not just the fact but what the eth­i­cal impli­ca­tions that stems from that fact—that that’s cen­tral in fact to the entire effort to pro­duce a whole new series of algo­rithms, the premise of which are the lim­i­ta­tions of human abil­i­ty, and the way to get com­put­ers to help us do that. 

Now, haunt­ing our dis­cus­sion today… So, what I’m inter­est­ed in is that we, and many of the peo­ple that were inter­est­ed to dis­cuss the pos­si­bil­i­ty of gov­ern­ing, are simul­ta­ne­ous­ly oper­at­ing in a con­di­tion of the inscrutabil­i­ty of that which we want to know. So, haunt­ing much of our dis­cus­sion of algo­rithms is I think a con­cern that we would become wor­ried when algo­rithms gov­ern us and not us them. 

And this isn’t sort of a sim­ple ver­sion where there’s a grand machine Matrix-like that’s mak­ing us think some­thing, or for philoso­phers the Cartesian demon, but rather a dis­per­sion of algo­rithms that are even hard­er to pin down. To use an algo­rithm well autonomous­ly is to lay down the law of using it. It is to know in some fun­da­men­tal way. 

Now, Lucas chal­lenges this analy­sis I think in deeply fu—in at least three ways. First, there’s no sim­ply know­ing an algo­rithm. Second, there’s no sort of pre-given human per­son­hood or sub­ject that is the ori­gin that we’re going to use in judg­ing algo­rithms. So we need to look at peo­ple dif­fer­ent­ly and the process of them. And then third­ly, there’s a real­ly grave dan­ger when we’re doing either sort of high-faluting aca­d­e­m­ic work look­ing at the effects of algo­rithms, or think­ing about con­crete eth­i­cal and polit­i­cal, or indeed cod­ing chal­lenges to exam­in­ing them. 

So, to begin look­ing at this, I want to begin with— It’s some­thing that’s an old chest­nut among this com­mu­ni­ty. I most­ly spend my time with bor­ing his­to­ri­ans so they don’t think this is so unex­cit­ing. But in the very first PageRank paper there’s a remark­able thing, which is of course that it’s always about per­son­al­ized search. It was always going to have it. Brin and Page wrote, Personalized page ranks may have a num­ber of appli­ca­tions…” Indeed. “…includ­ing per­son­al search engines. These search engines could save users a great deal of trou­ble by effi­cient­ly guess­ing a large part of their inter­ests giv­en sim­ple input such as their book­marks or home page.”

Now, from this sim­ple begin­ning of course some­thing more search­ing arose. Fundamental to defens­es of the data min­ers of users is a claim indeed that the cog­ni­tive limitations—that’s our inabil­i­ty to think well—of human beings require us both to data mine to learn things, but as cen­tral­ly to be data mined. In this vision, data min­ing helps us to become free. Because the space of poten­tial things we might go and learn about is sim­ply too large, if it were acces­si­ble. In fact, being data mined allows us in this argu­ment to move from a sort of for­mal dec­la­ra­tion that we are free to go and learn any­thing, towards an actu­al­iza­tion, because we are being helped. Something that is judg­ing us is help­ing us to over­come our own cog­ni­tive limitations. 

Now this kind of argu­ment is com­ple­ment­ed by the sorts of argu­ments that the mas­sive data bro­kers make. And there’s a remark­able set of— How many min­utes? Three. Okay, I’ll be quick.

So the data bro­kers quite remark­ably make a sim­i­lar argu­ment. One of them says, for many pop­u­la­tions for whom online ser­vices are made free, infor­ma­tion tru­ly is a direct con­duit to Liberty.” So there’s a dou­ble enabling of free­dom, that is, that allows us to become the peo­ple we want to be. To actu­al­ize our­selves. To become who we real­ly are. And what’s inter­est­ing is Lucas is giv­ing us an account, using the per­for­ma­tive the­o­ry, of becom­ing in which there isn’t an essence. But many of the peo­ple who are inter­est­ed in gov­ern­ing are pro­found­ly inter­est­ed in pro­vid­ing ser­vices that are about becom­ing who we real­ly are. And I think that shared and dif­fer­ent ontol­ogy is well worth think­ing about. 

Okay. I want to end just on this ques­tion of gov­ern­ing with­out knowl­edge. And we keep com­ing back to this when we talk about trans­paren­cy. We’ve talked about the prob­lems of trans­paren­cy both epis­te­mo­log­i­cal (we can’t real­ly know these sorts of things), as well as things like trade secrets. We can’t gov­ern through knowl­edge, prop­er­ly speak­ing. Even if many algo­rithms are trade secrets, Lucas and oth­ers have remind­ed us near­ly all would not be sur­veil­l­able by human beings, even if we had access to their source code. We have to begin what­ev­er process from this fun­da­men­tal lack of knowl­edge. We need to start from the same epis­te­mo­log­i­cal place that many of the pro­duc­ers of algo­rithms do. 

And so I think, curi­ous­ly enough, at the moment that we’re critical—as I think we right­ly are—of prox­ies like Turnitin, we’re in des­per­ate need of prox­ies in think­ing about how to gov­ern the kinds of algo­rithms that wor­ry us. We can­not know them. Even if some­one hand­ed them to us, we could­n’t know them. And so I think that’s one of the sort of things we real­ly need to know. We real­ly need to pro­duce prox­ies. Now I sus­pect, and here I’ll end, that this is of course going to cause gam­ing the sys­tems. But I won­der if mutually-assured gam­ing isn’t one of the best things we can do if we’re inter­est­ed in gov­ern­ing. And those who are inter­est­ed in gov­ern­ing us…we have a shared inter­est in actu­al­ly both gam­ing the sys­tem. Okay, I’ll end there.

Further Reference

Jones’ response paper