On the Exculpatory Moral Economy of AI-DSS

Abstract

This article analyzes the foundational concepts underpinning the European Union’s governance of artificial intelligence decision support systems (AI-DSSs): “trustworthy AI” and “human oversight.” Evidence from behavioral psychology, computer science, and ethics and legal scholarship demonstrates that automation bias, selective adherence, and AI hypersuasion systematically undermine meaningful human agency (MHA). Furthermore, the anthropomorphic discourse of trustworthiness obscures moral responsibility rather than ensuring accountability. This situation raises a significant question left unresolved in the literature: why does the oversight paradigm persist and even expand despite its well-documented insufficiency? To address this, the article applies the moral economy framework, as developed by E. P. Thompson and further elaborated by Bonini and Treré, to the context of AI-DSS deployment. It delineates four key positions: deployers, subjects, resistors, and the conscripted human-in-the-loop. The analysis identifies the dominant value system in this field as an exculpatory moral economy, in which protective norms are fulfilled through performance rather than genuine protection, resulting in the receipt of trust and the granting of absolution. From this perspective, the failure of oversight is not a flaw to be corrected but an intended outcome of the arrangement. The article advocates for a strong presumption against deploying AI-DSSs in sensitive areas of societal governance. Where deployment continues, it proposes two legally grounded avenues for organized societal resistance to uncritical AI-DSS deployment: promoting moral economy-aware critical AI literacy, which furthers skepticism rather than trust, and establishing collective MHA councils, rooted in the AI Act and the precautionary principle.