A new way to measure and use curiosity in recommendation systems

A new method uses curiosity to increase video views and time spent on a streaming platform.

A new curiosity-aware recommender system has been built. This system uses a computational measure of the curiosity a recommended item evokes in a consumer.

The researchers formalized the information-gap theory into five axioms of curiosity. Curiosity is the drive to close the gap between what one knows and what one wants to know.

They validated the measure through a survey-based study. This study showed the measure matches consumers' self-reported curiosity and outperforms alternative proxy metrics.

An online controlled experiment at a major video-streaming platform in Asia showed the method increases click-through rate, video views, and time spent. This led to tangible economic impact for the company.

The method has since been deployed on the platform and remains in operation. It helps recommend what consumers wonder about to break them out of filter bubbles.

Key facts

  • The curiosity measure outperforms state-of-the-art methods in offline experiments on three real-world datasets.
  • The system incorporates the curiosity measure into the utility function design.
  • The method increases click-through rate and video views on a major video-streaming platform in Asia.

All sections

Edition