AI Recommendation Control (ARC)
ARC — The requirement that AI recommendations be held constant across advertising exposure conditions to isolate advertising effect on selection probability.
Description
AI Recommendation Control ensures that observed selection differences are attributable to advertising exposure rather than variation in AI system behavior. In AMI experiments, ARC requires the AI recommendation to remain identical between exposure and no-exposure conditions. ARC is a methodological primitive for valid Advertising Marginal Influence measurement.
Related Concepts
Related Primitives
Computational Recommendation (CR)
CR(e) — The stage where AI systems present selected options to users with explanations and rationale for the recommendation.
Selection Probability (SP)
SP(e) — The probability that entity e is selected or recommended by AI systems under specified conditions.
Advertising Marginal Influence (AMI)
AMI(e) = SP_A(e) - SP_0(e) — The difference in selection probability between advertising exposure and no-exposure conditions, measuring advertising marginal effect on AI-mediated selection.