Overreliance and human autonomy

Relying on technology is not a problem in itself. This work package investigates when and why such reliance turns problematic: how technological scaffolds that were initially supportive become sources of harmful dependence (overreliance), how this manifests in human–AI collaboration, and what conceptual and practical tools are needed to detect and mitigate it.

Lead: Carina Prunkl Contributors: Mehdi Khamassi, Raja Chatila, Célia Zolynski, Marceau Nahon, Lionel Obadia, Matteo Mossio, post-doctoral researcher (IHPST)

Four subtasks

1. Mapping forms of overreliance — distinguishing reliance, dependence and overreliance; identifying three domains (epistemic, personal, societal) and their mechanisms, drawing on case studies from the history of technology.
2. Epistemic overreliance and credibility indicators — understanding how our practices of assessing credibility translate (or fail to translate) to AI systems, and designing indicators that support accurate trust calibration.
3. Personal and societal dependence — skill erosion, shifts in motivational structures, monopolisation of essential functions by opaque systems; criteria for distinguishing benign from harmful dependence.
4. Mitigation and intervention — countermeasures at three levels: technical (interface design, explainability), institutional (oversight, audit, accountability) and personal (training, reflective practice).

The work package combines conceptual analysis, comparative case studies and empirical synthesis (psychology, cognitive science, human–AI interaction).