Sensibilities and Forms of Delegation to Generative AI: An Analysis of Problematized Narratives

Abstract

The corpus studied in this article brings together 33 narratives where users question their own use of generative AI (IAg), and in particular conversational AI (cIA), whether in an academic, professional or personal contexts. It focuses on participants' experiences and the forms of delegation to cAI – ranging from substitution to cooperation, and even refusal. The analysis is grounded in three concepts: the share of self engaged in or delegated through the interaction (Spach, 2025); registers of activity, which capture the broad orientations that shape interactions with cAI; and sensitivity to (Récopé et al., 2013). These concepts are applied to the narrative analysis using a mixed coding approach (Paillé and Mucchielli, 2016). The results reveal four dominant activity registers – operational economy, epistemic vigilance, subjectivation and self-preservation, and axiological coherence – and six distinct sensitivities to that explain why the same task performed with the cAI can be experienced differently from one person to another. The study shows that epistemic agency is not a stable property of individuals, but rather a potentiality that is activated or inhibited depending on the dominant sensitivity to and the configuration of the situation. The study underscores the importance of designing educational interventions that activate sensitivities conducive to critical reflection and vigilance.