Stakes, Context, and AI How AI Challenges Human Ways of Knowing

Activiteit: Conferentie of presentatieMondelinge wetenschappelijke presentatie

Beschrijving

This talk examines how AI systems operate as "natural evidentialists" that base outputs solely on training data patterns. I exploit an example of the use of Augmented Reality technology in the military. While these systems can be fine-tuned with top-down corrections for context and stakes, they retain a problematic "evidence-overlay" structure that treats evidential factors as primary and moral or pragmatic considerations as secondary disclaimers. This flattens complex human judgment processes into algorithmic rule-application, risking both moral and epistemic deskilling as we become less practiced at exercising virtue-driven reasoning in contexts where stakes, values, and non-evidential factors matter.
Periode19 sep. 2025
EvenementstitelETHICOMP 2025: 22nd International Conference on the Ethical and Social Impacts of ICT
EvenementstypeCongres
LocatieLisbon, PortugalToon op kaart
Mate van erkenningInternationaal