Developing AI for Weather Prediction

: Ethics of Design and Anxieties about Automation at the US Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography

Abstract

The question of how professional and lay communities develop trust in new technologies, and automation in particular, has been a matter of lively debate. As a charismatic technology, artificial intelligence (A.I.) has been a common topic of these debates. This paper presents a case study of how the discourses and principles of ethics of technology development—specifically, of A.I.— were mobilized to form trust among actors in the fields of computer science, risk communication, and weather forecasting. My analysis draws on sociology of expertise and the literature on ethics of A.I. to ask: how emerging networks of expertise use ethics to overcome mistrust in technology? And, what role does the institutionalization of those networks play in the process of trust formation? I situate this discussion on the NSF Institute for Research on Trustworthy A.I. The Institute is positioned as a mediating organization with the goal of increasing trust in this technology primarily the weather forecasting community, but also among the public. I show that first, to better understand how scientific and professional fields react to increased automation it is crucial to unpack the historical backdrop of how the professional identity of those experts has been shaped by a relationship with computer-supported modeling. To this end, I situate the discussion in the long-standing tensions between computer modelling and tacit knowledge in weather forecasting. Second, I argue that the means of establishing trust in A.I. propagated by the actors in the paper, which pair norms of explainability to sensitivity to professional intuitions and domain-specific conventions, rely on a series of “mutual orientations” (Edwards, 1996). I mobilize the concept of “mutual orientations” to describe the work of tailoring the ethics of A.I. to the specific requirements of weather sciences, but also to the vision of a national strategy of investment in this technology.