Are conscious machines valuers?
AI & SOCIETY May 1, 2026 Jan Henrik Wasserziehr
Even if artificial systems achieve consciousness through computational functionalism, they are unlikely to experience valenced states—feelings of pleasure or pain—because they lack a non-derivative goal like self-preservation that grounds value for living organisms. This creates a value grounding problem: without such a disposition, there is no basis for some states to be objectively better or worse for the system itself. The paper examines four possible routes to artificial valence—designer-independent goals, reinforcement learning, rational evaluation, and hallucinations—and concludes that none resolve this problem. Consequently, recent proposals for artificial consciousness probably do not yield sentience.