Superficialism About Belief, and How We Will Decide That Robots "Believe"

Abstrakt

Superficialism about property X treats the possession of property X as determined entirely by superficial as opposed to deep facts.  Belief should be understood superficially.  Facts about belief are determined entirely by facts about actual and potential behavior and conscious experience ? plus transitional cognitive states ultimately understood in terms of actual and potential behavior and conscious experience.  Superficialism about belief excels on both intuitive and pragmatic grounds, compared to accounts of belief in terms of deep cognitive or neural architecture.  Behavior-focused superficialism suggests that robots and Large Language Models may soon (perhaps already do) believe.  If consciousness is also essential to belief, matters might soon become more complicated, if it becomes reasonable to wonder whether some of our most advanced AI systems are conscious.  Regardless, it will be practical to describe some such systems as having what I?ll call ?belief*? ? belief shorn of commitment to any conscious aspect ? and people might forgivably forget to pronounce the asterisk.  Krzysztof Posłajko should welcome this manner of thinking, though it needn?t be as ?antirealist? as he suggests.