Human-in-the-Loop for Artificial Intelligence

Le résumé fourni par la source

With the introduction of AI agents, more data, faster computer processing hardware, and more interconnected third party developers, the ability for humans to manage or interfere with AI autonomous decisions is decreasing. Human agency is a key criteria for accountability. Previously, for oversight, agency was provided broadly by the concept of human-in-the-loop (HITL), however it is unclear if this is still relevant. This chapter briefly outline the concept of HITL which originated from a military context. In this paper we argue that HITL is difficult to implement, and it is often used as AI Ethics Washing, where the term is used to protect AI companies and stakeholders and blame the users. We introduce a framework for HITL where it moves away from being discretionary where AI and human agency are dependent on the transparency accountability of the power, trust and complexity to prevent AI Ethics Washing.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.