Perceptions of agency : untangling the knotty web of Al

Abstract

Artificial intelligence systems (Al) have become a ubiquitous part of modern life. Yet their complexity has prevented a concrete conceptualization that correctly map the web of human actors and computational processes involved in Al. This opaque representation of AI poses questions for accountability and governance, such as who is responsible when an Al makes a moral transgression? This thesis takes a discursive and empirical approach to reifying Al as a specific network of human actors with real world outcomes. It explores the phenomenon of anthropormophization, by which Al is endowed with human-like characteristics, and shows how the extent to which a Al system is anthropomorphized can affect the attribution of responsibility to human actors. This thesis does not offer a normative suggestion for whom society should blame when Al make moral transgressions, but rather offers a view into human folk intuitions.