Skip to main content

Intentionality

An action can be directed at an object, a participant can report what they are thinking about, and a model can carry a variable labelled “target.” These descriptions have different observation conditions. A model of intentionality should predict what changes when the target, cue or prospective outcome is varied independently.

Distinguishing targets from cues​

Show the same cue with two objects and change which object will yield a stated outcome. Transfer to a new cue or a new route can distinguish tracking the object from repeating a learned movement. A report about the object is an additional record. Matching one cue–action sequence cannot identify all three mechanisms.

What closure under composition gives​

Suppose objects carry states and a specified auxiliary readout ρE\rho_E. If one defines admissible arrows to be CPTP maps preserving the inequality r(B)≥r(A)r(B)\ge r(A) for r(A)=rank⁡ρE(A)r(A)=\operatorname{rank}\rho_E^{(A)}, identities are admissible and two successive such arrows are admissible. This follows from CPTP composition and transitivity of the rank inequality. The identity channel already satisfies the condition for every object. Thus the condition alone distinguishes neither a target nor a particular act of thinking. On a joint system a partial trace needs an actual tensor factorization; a labelled coordinate of C7\mathbb C^7 supplies none. Quantum instruments

What a report can establish​

Record which object the participant names or indicates, when the question is asked, and whether the response predicts later choices for a changed target. Reports can constrain an account of what an activity is about. A claim about unreported experience needs an independent discriminator. Phenomenal distinctions