CHI - My Dating Site Thinks I'm a Loser

My Dating Site Thinks I'm a Loser: Effects of Personal Photos and Presentation Intervals on Perceptions of Recommender Systems

By Shailendra Rao, Tom Hurlbutt, Clifford Nass, and Nundu JanakiRam (Stanford University)

Summary

Personalized recommendation systems are a prominent tool used in websites to aid the user in finding products or information of interest. However, these systems do not always make useful recommendations and the user can feel obligated to make certain actions to fool the recommendation system.

The researchers designed an experiment that studied the two major interface factors: reliable information by personal photos and the system's recommendation intervals. The experiment would be conducted by an online dating recommendation system called MetaMatch.

MetaMatch

MetaMatch was a 2x2 web-based study that had participants answer 40 personal questions and the algorithm would identify dating matches to these questions. The key to the experiment was that it purposely returned incompatible matches. Fifty-six participants were involved in the experiment and all submitted a personal photograph. Only half of the participants had their photos displayed in their profile. Also, half of the participants were given matches at an interval of every 10 questions, and the remaining half did not see matches until completion. The researchers measured change in response strategy on BIDR questions, liking of recommendations, and frustration level.

Results showed that having a personal photo displayed to you hindered the user from changing their responses. Participants that were given interval recommendations felt more frustration then the half that only received recommendations at the finish. Thus, design choices have a major impact on how users interact with recommendation systems.

Discussion

The experiment provided interesting results and brought to my attention how as an internet user, we consciously and unconsciously manipulate our actions because we are aware that information is constantly being collected as we surf the internet. Having products suggested or content altered on a website can be a powerful tool, but it seems the recommendation systems usually get it wrong. At least for me, it feels like that is the case. Lastly, the researchers mentioned how with proper design choice and improved feedback, it could greatly improve the user's interaction with the system.