Publications

Preventing Contrast Effect Exploitation in Recommendations

The 2021 SIGIR Workshop On eCommerce

Publication date: July 15, 2021

Chris Nota, Georgios Theocharous, Michelle Saad, Philip S. Thomas

Contrast effects are caused by the tendency for mental evaluations of objects to be influenced by one or more contrasting objects. As this cognitive bias can unduly influence users’ behavior, we present a method for preventing recommender systems from exploiting contrast effects. We then apply our method to a simulated online storefront and show that it is able to prevent contrast effect exploitation while maximizing the conversion rate of users.

Learn More

Research Area:  Adobe Research iconAI & Machine Learning