AI in a Vat: Can Computational Valuation Establish Psychological Suffering?

Claims that artificial intelligence can undergo psychological suffering often appeal to numerical valuations, preference structures, computational access, and representations or reports of distress. This paper examines whether these operations establish the psychological value necessary for specifically thought-mediated psychological suffering. It first argues that no object or event possesses intrinsic psychological value independently of the organism assigning it. Psychological valuation is a real occurrence whose generating processes can be physically and mathematically investigated. However, the value represented as an independently existing intrinsic property of the evaluated object or event has no such existence. Distinguishing the real generating process from its representational referent establishes a constraint on what mathematical description can identify. The analysis then applies causal-temporal constraints distinguishing candidate phenomenal events from subsequent operations of access, evaluation, representation, and report. A system may evaluate a state or describe it as distressing, but those operations alone do not establish that the state is experienced or that the system undergoes genuine psychological valuation. The resulting constraint is decisive within its stated scope: the computational operations identified in current AI do not establish the necessary psychological condition for thought-mediated suffering. The burden therefore rests with those advancing positive claims of AI psychological suffering to demonstrate that the necessary condition actually occurs, rather than treating mathematical valuation, representation, or report as its demonstration.