Possible Theoretical Approach
Treating MTurk and HITs and Yep, AAI as cultural techniques in the style of ANT we want to trace how the line between the human and the nonhuman is drawn by the articulation of HITs within MTurk. This is why the set-up and framework—the infrastructure of the project—matters so much. We start by going through the database and then bring in people to talk through what we found—bringing people and readings to help us describe what we are seeing. This is a way both to bring people on and give us a focal point for the conversation: give them something they can do through their own work.
Our approach is something of a hybrid between actor-network theory and cultural techniques. Cultural techniques is generally an historical approach that marks the chains of operations that have produced the concepts that, retroactively, seemingly caused them. For cultural techniques, the binary of inside/outside is produced by the door, the distinction between human and animal by the fence. ANT, which Siegert draws on, tends, by our reading, to be a more present-oriented research practice. Because ANT reassembles it is always kairotically assembling now, in the present. With this hybrid, for which no reasonable acronym exists, we want to trace the chain of contemporaneous operations that are currently conceptualizing as it produces the distinction between human and machine, both of which are already the outcomes of cultural techniques—hence, the need for ANT in concert with CT. We cannot use “the human” to assess or evaluate the MTurk project ahead of time, as the human is precisely what is at stake (always) by its formation through a set of cultural techniques operating presently. We are thus tracing a cultural technique that is itself still being composed—both human and cultural technique are thus being reassembled within this project.
This approach renders our project somewhat conceptually complex and, in practice, mind-numbingly simple. MTurk, in short, makes it pretty easy for a couple of ants to follow the actors, which is already revealing. It is already databased and organized around activity. And the arrangement of the database, its articulation, is precisely what we wish to trace and learn from. MTurk, using the language of CT, isn’t organized around a conception of the human but rather a set of needs—labor requirements—the meeting of which generates a more-or-less stable and recognizable, we wager, conception of the human as distinct from the machine.
In line with cultural techniques, whatever the human is, it’s been carved out of and by distinctions: the barn doors that keep the nonhuman out. What is potentially new here, is that MTurk articulates the human not with (barn)doors but with a cabinet that allows a kind of hidden slippage between the human and the machine that allows a kind of hidden slippage between the human and the machine. The original, as with the contemporary, mechanical turk is decidedly not not a machine and not not a human.
We don’t lose but gain critical power here. There’s a supposed risk in uncritically wading into the detritus of Silicon Valley. Going in unarmed, as it were, risks turning ourselves over to the fantasies of libertarian tech bros. Surely we need to know what a human is so that we might judge those whose code calls the tune. We think it’s worthwhile to follow them around a bit—to follow the labor and the discourse and even the money: the cultural techniques. How is the human being carved out in the mechanical turk? What do HITS make of the human?
Methodologically, we’re (likely) going to be using AI collaboration to compose this research; We should point out the nature of this cooperation. It’s possible to see automation as a winnowing of the Platonic human province, or as a largely economic displacement of humans by machines. But that can cut both ways. We can use its tools to understand all of us.