Epistemology Gives a Future to Complementarity in Human-AI Interactions
The paper argues that human‑AI complementarity, the idea that a human supported by AI can outperform either alone, faces theoretical challenges such as lack of precise anchoring and reliance on post‑hoc accuracy. By applying epistemology and computational reliabilism, the authors reframe complementarity as evidence of a reliable epistemic process, linking it to other reliability indicators and socio‑technical practices. They propose design and governance recommendations, including a minimal reporting checklist and measures of efficient complementarity.
The paper argues that human‑AI complementarity, the idea that a human supported by AI can outperform either alone, faces theoretical challenges such as lack of precise anchoring and reliance on post‑hoc accuracy. By applying epistemology and computational reliabilism, the authors reframe complementarity as evidence of a reliable epistemic process, linking it to other reliability indicators and socio‑technical practices. They propose design and governance recommendations, including a minimal reporting checklist and measures of efficient complementarity.
This work informs Swiss AI policy by providing a framework for assessing the reliability of AI‑supported decision‑making, which is essential for regulatory compliance and responsible deployment in Swiss industry and healthcare.
Original source record. Open the original record to verify the underlying announcement.