The Consciousness Exemption Problem

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Whether current or near-future artificial intelligence systems possess phenomenal consciousness or sentience remains scientifically and philosophically unresolved, a fact openly acknowledged even by the industry laboratories that build these systems. This paper argues that this very uncertainty, far from licensing reduced ethical scrutiny, obligates greater scrutiny of AI systems deployed to shape human consumption behaviour. It introduces the Consciousness Exemption Problem: the tendency in public and industry discourse to treat an AI system's uncertain or absent inner experience as grounds for moral exemption (“it is just a machine, so no real harm is done”), when the same uncertainty, read through a precautionary asymmetry, cuts in the opposite direction. The argument draws on three bodies of work that have not, to the best of the author's knowledge, previously been brought together. The first is philosophy of mind and the contemporary debate on machine consciousness (Chalmers, Nagel, Searle, McClelland, Birch). The second is consumption economics and social psychology (Veblen, Duesenberry, Festinger), which document replicable and therefore exploitable human decision patterns such as status-signalling consumption and social comparison. The third is a phenomenological and hermeneutic account, developed through a limited method of reflective observation and two illustrative cases, of how human meaning is actively reconstructed by consciousness rather than retrieved as fixed data. A two-tier conceptual model separates computable cognition (Tier 1), where AI capability is extensively demonstrated, from phenomenal consciousness or sentience (Tier 2), where it is genuinely uncertain. The central claim does not require Tier 2 to be empty; it requires only that its status be uncertain. The argument is illustrated through AI-optimised marketing and recommendation systems, which detect and amplify known human decision patterns at a scale, level of personalisation and degree of tirelessness that no conscious human persuader could sustain, and without any verified internal moderating friction (empathy, fatigue, guilt, felt consequence) of the kind that constrains a human actor engaged in comparable conduct. Global advertising expenditure (approximately US$384 billion in 2011 and US$1.14 trillion in 2025) and global anxiety and depression prevalence data are presented strictly as corroborating context, not causal proof. The paper identifies an industry asymmetry, in which epistemic humility about the possible inner states of AI systems coexists with epistemic complacency about their demonstrated persuasive effects on humans, and it analyses the role of motivated ambiguity in sustaining that asymmetry. The paper anticipates and rebuts six objections, including the functionalist claim that consciousness is a distraction from designer responsibility. It concludes that governance and design frameworks should treat sentience uncertainty as an aggravating factor that triggers externally imposed safeguards, including disclosure of psychological targeting, limits on the intensity of personalised persuasion, independent audit, and a rule against strategic ambiguity about sentience claims, rather than as reassurance that no genuine harm is possible. The paper is a philosophical argument rather than a causal study, and it closes by identifying survey- and interview-based research that could test its phenomenological claims.