AI and the Iterable Epistopics of Risk

TL;DR

This paper explores how AI risk management relies on everyday practices rather than solely on formal frameworks, highlighting the importance of situated knowledge and iterative topics in understanding and addressing AI risks.

Contribution

It introduces the concept of iterable epistopics of risk, emphasizing the role of situated practices in AI risk management and proposing a new interdisciplinary approach.

Findings

Risk management depends on mundane, situated practices.

Iterative epistopics reveal how practitioners understand and respond to AI risk.

The approach informs development, regulation, and future frameworks.

Abstract

Abstract. The risks AI presents to society are broadly understood to be manageable through general calculus, i.e., general frameworks designed to enable those involved in the development of AI to apprehend and manage risk, such as AI impact assessments, ethical frameworks, emerging international standards, and regulations. This paper elaborates how risk is apprehended and managed by a regulator, developer and cyber-security expert. It reveals that risk and risk management is dependent on mundane situated practices not encapsulated in general calculus. Situated practice surfaces iterable epistopics, revealing how those involved in the development of AI know and subsequently respond to risk and uncover major challenges in their work. The ongoing discovery and elaboration of epistopics of risk in AI a) furnishes a potential program of interdisciplinary inquiry, b) provides AI developers with a means of apprehending risk, and c) informs the ongoing evolution of general calculus.