User-generated reviews of hospitality services and retail outlets (such as restaurants, hotels etc) are highly effective in increasing revenue through social-proof. Studies have shown that recommender systems can earn 10% - 30% of additional revenue for a company. However, recommendations in local outlets can easily be falsely acquired by service providers, by ‘bribing’ customers in exchange for good reviews. Comparator sites and recommender engines stand to lose, as false recommendations can lead to dissatisfied customers, hence user-churn over time. Traditionally, recommender systems rely on ‘Objective Rating’ (or ‘O-rating’): individual evaluations are aggregated into a single figure, which is seen by, and thus influences, every potential customer. The O-rating is highly prone to bribery, and is thus sub-optimal for comparator sites and recommender engines.
Game theory and artificial intelligence researchers in computer science have devised a novel ‘Personalised Rating’ (or ‘P-rating’), which is fully bribery-proof in instances where the service provider has a cost in bribing an individual. Using network-based voting and mechanism design, the P-rating model leverages social circles within the network to augment the quality and reliability of the reviews. Analysis has shown that the P-rating is more bribery-proof than the standard O-rating in general, and particularly when the service provider must bear costs for bribery. Adoption of the P-rating should hence lead to greater satisfaction in customers over time.
Bribery-proofed recommender systems for physical outlets.
Early-stage.
Recommender systems.
Alex Garcia
