(Re)designing AI for Diverse Disciplines

The (re)emergence of artificial intelligence in science raises questions concerned with how fields can optimally benefit from current and imagined future capabilities of the technology. One approach that has been proposed (and heavily funded) is widescale training in AI tools and techniques as a means of both upskilling researchers and further diffusing the technology into various disciplines that are perceived as not maximizing gains on AI’s potential. Yet, very little attention has been paid to the enormous epistemic and normative heterogeneity that serves in part to distinguish the various scientific and scholastic disciplines and the role that such distinctions play in determining the questions and methods of inquiry that fields take up. 

One worry, then, is that, while the epistemic and normative commitments of some disciplines put them in a position to readily adopt and adapt to AI in a relatively straightforward manner, others must either modify their fundamental standards, questions, and approaches so as to accommodate the use of AI or carry on without the aid of these breakthrough technologies. 

We propose a workshop to ask and begin to answer big questions concerning disciplinary factors that drive successful engagement with AI, as well as to develop and extend collaborations across disciplines and institutions that realize locally optimal rather than maximal engagement. In this, we aim to transform the dialogue around these topics by inverting the narrative and putting disciplines first. Rather than ask ‘how might one adapt the commitments of a particular discipline to the capabilities of AI?’, we ask ‘how might we adapt the capabilities of AI to optimize of diverse disciplines?’. In this way, we hope to set an early agenda to responsibly steer the evolution of the relationship between AI and the disciplines.