By speaker > Hsiao Ming-Yuan

Discussions of artificial moral agency often maintain that current AI systems

cannot be morally responsible because they lack capacities associated with

responsible agency, such as autonomy, normative self-governance, and

meaningful control. Although rarely framed in terms of the Principle of

Alternative Possibilities (PAP), this view typically assumes that responsibility

requires access to genuine alternatives. I argue that we lack the epistemic

resources to determine whether any agent, human or artificial, possesses such

alternatives. On an epistemic interpretation of Frankfurt-style cases, the central

issue is not whether PAP is false, but whether we can know whether an agent

could have done otherwise. This uncertainty extends even to paradigmatic

human agents. If we treat humans as responsible under this opacity,

consistency prevents us from excluding AI on the same basis. I call this

epistemic parity. Responsibility practices should be guided not by unverifiable

metaphysical assumptions, but by what can be known.