Paper short abstract:

Once STS invoked microbes that disrupted how things were done. So does AI in its anthropomorphism and as an interpretive relation to nature. "To let nature back in” should, again, be about the nature-cultures of AI and how they invoke STS as a specific relation to political and democratic theory.

Paper long abstract:

The fascination with AI recalls how STS pointed to the laboratory as the principle setting where to understand the persuasiveness, the ability to circulate and the disruptiveness of science and technology. Now it is ChatGPT that is breaking out of a controlled environment, showing its ability to unsettle established ways of thinking and acting.

Like laboratory studies revolve around an interpretive relation to life forms and nature, AI is consistently framed in anthropomorphic terms. AI is seen in terms of its ability to train, learn and think, while there is something natural about language processing. Consequently the approach could still be the same: establish proximity to cases but remain detached from hype and speculation. This holds even though the sociology of expectations increasingly can be extended to STS itself: having raised expectations about the ability to meaningfully redesign and govern AI. The tables turn.

What makes a good case responsible rather than disruptive, open rather than opaque, trustworthy rather than persuasive, human-centric rather than being excessive about non-humans in environmental terms? Is it once more about “letting nature back in”, perhaps with a bigger variety of settings? Or is AI the moment that STS recalls that its “nature-cultures” were once about social contract theory (mostly Hobbes), about constitutionalism, and a democratic theory that revolves around the ability to experiment?

This paper affirms that AI needs STS, as the billions pour into OpenAI, ConstitutionalAI and similar entities. It also argues that now is the time to return to conventional political theory.