Feedback must support causal inference
Aliases: causal feedback · action-outcome mapping · informative feedback · feedback timing
What it is
Feedback for causal inference means the game's feedback on player actions must let the player judge "which of my actions caused this outcome." Not all feedback meets that standard: delayed feedback, feedback where multiple candidate causes coexist, or feedback with no perceptible link between action and outcome cannot support inference, and players learn no rule from it.
Why it happens
Causal inference relies on three temporal properties: contiguity (feedback follows the action closely—beyond about one second the causal link blurs), consistency (the same kind of action produces the same kind of result, or no stable action-to-outcome mapping forms), and exclusivity (when the result occurs, no other unexplained change happens simultaneously, or attribution splits across candidates). Each feedback channel has a speed advantage—haptic is fastest (near-instantaneous), visual next, audio best for marking state changes. Channel choice should match the causal chain length: instant hits use haptics, delayed explosions use visual preparatory signals (flashing lights, smoke) to mark the causal source ahead of time.
Where it stops holding
Contiguity does not mean instantaneous. Some mechanics are inherently delayed in effect (poison, timed bombs, strategy turns); the feedback system must bridge the delay with preparatory signals rather than pretend the delay is absent. In networked multiplayer, network latency naturally inserts itself between action and outcome; causal perception then needs client-side prediction or lag compensation to hold. Exclusivity is hardest in complex systems—when several systems act at once, players may attribute the outcome to the wrong system and form an "effective but wrong" strategic belief that only collapses when a counterexample appears.
Applying it
- Audit each core operation against the three properties: action-to-feedback latency, feedback consistency across instances of the same action, and whether any unexplained concurrent change accompanies the outcome.
- Design preparatory signals for delayed consequences (visual flash, audio cue) so players can track the causal chain across the delay window.
- Verification: record players' spoken explanations of "what caused what" and compare them to the actual rules. Misattributed rules mean one of the three properties failed; locate and repair it.