Reseach Discussion 3

Do recommender systems reinforce human bias?

The Cambridge dictionary defines ‘bias’ as “the action of supporting or opposing a particular person or thing in an unfair way, because of allowing personal opinions to influence your judgment” and “the fact of preferring a particular subject or thing.” In that sense, we all have our preferences and inclinations, but the real issue is how do we separate them from unfairness. How can we be objective about something be it a product, an idea, a movie without the influence of the group. I find this topic to be particularly riveting since technically recommender systems are built to measure ‘bias’. Moreover, I do believe that when building recommender systems, we have to keep bias in mind. Completely removing bias from recommender systems is not right if bias exists organically within our society. In their article, Krishnan et.al. discuss the affect of social biases on recommender systems and the machine leaning methods of reducing those biases. “We propose a methodology to 1) learn, 2) analyze, and 3) mitigate the effect of social influence bias in recommender systems”, the authors write in the abstract. In the Learning stage a dataset is obtained by allowing participants to rate items twice, once on their own and another time after seeing other participants’ ratings. In the Analysis stage, with the help of the non-parametric significance test the presence of bias is assessed. If there is bias then the Mitigation stage occurs “where mathematical models are constructed from this data using polynomial regression and the Bayesian Information Criterion (BIC) and then inverted to produce a filter that can reduce the effect of social influence bias.” I do believe that it is impossible to get away from bias when we are talking about recommender systems for technically the a well-built system implements bias in order to make good recommendations.

Reference: Krishnan, S., Patel, J., Franklin, M. J., & Goldberg, K. (2014). A methodology for learning, analyzing, and mitigating social influence bias in recommender systems. Proceedings of the 8th ACM Conference on Recommender Systems - RecSys 14. doi: 10.1145/2645710.2645740