Learning From Social Interactions: Personalized Pricing and Buyer Manipulation

A new study published in IEEE Transactions on Mobile Computing provides a formal game-theoretic analysis of the escalating arms race between AI-driven personalized pricing and consumer counter-strategies. Researchers from The Chinese University of Hong Kong, Shenzhen, and the University of Macau modeled how platforms like Amazon infer a new buyer's private preferences and willingness to pay by analyzing the purchase histories of their social connections—a practice grounded in the sociological principle of homophily. Their parsimonious model captures a double-layered information asymmetry, where the seller lacks perfect knowledge of individual preferences but can learn from correlated social data, while buyers can strategically alter their visible social interactions to mislead the algorithm.

The analysis yields counterintuitive results. It finds that only buyers with a high intrinsic preference for a product engage in costly social signal manipulation, such as hiding friendships or curating their network, to appear less correlated with high-spending peers. Surprisingly, this strategic evasion often leads to worse financial outcomes for the manipulators themselves. Furthermore, the seller's revenue gain from implementing this social-data learning practice is significant and remains robust; buyer awareness and subsequent manipulation have only a marginal impact on the bottom line. Consequently, the research advises that in light of tightening data regulations, sellers benefit more from being transparent about their data access practices, aligning with informed-consent norms rather than operating opaquely.