Artificial Answerability
Whether, and how, algorithmic systems can be held to account for what they mediate.
Theory
Artificial answerability asks what it would mean for an automated system to be answerable — not merely explainable — for the outcomes it produces. Answerability is a relational, second-personal notion; it requires an addressee, a stake, and a capacity to receive challenge.
The concept distinguishes explanation (a technical output) from answerability (a moral and institutional relation). A system may be highly explainable and yet answerable to no one; conversely, an opaque system may be embedded in structures that hold specific persons answerable.
Artificial answerability provides criteria by which to evaluate proposed AI governance regimes: not by their transparency alone, but by the answerability relations they establish, sustain, or erode.
Key publications
- Foundational essay distinguishing explanation from answerability (Philosophy & Technology).
- Applied work on answerability in algorithmic healthcare (AI and Ethics).
- Theological engagement with answerability and the moral self (Studies in Christian Ethics).
A continually updated record is available through the publications page.
Current work
- A book-length treatment of answerability as a category for the philosophy of AI.
- Comparative analysis of proposed AI regulations through the lens of answerability.
- Interdisciplinary work with legal scholars on institutional forms of artificial answerability.