Analytics - Bases for Bias

Big data and data analytics are causing a massive change in the industry. Yet, we are just scratching the surface on analytics applications today. Some common areas where analytics has been applied successfully are : 1. To generate user recommendations, 2. Targeted ad campaigns and 3. Medical records analysis. When using analytics for any application, one must be wary of the bias introduced in the system, whether intentional or non-intentional.

Sites like Facebook, Netflix, AirBNB, etc use analytics and machine learning for generating user recommendations regarding which messages to display, which movies to see and suggested rentals. Facebook, Google and Yahoo may use to generate targeted ads based on the previous searches, etc. Medical records analysis can uncover geographic traits of a disease or the likelihood of diseases based on genetics. Bias may be defined as intentional or non-intentional skew that  can affect the outcome of a given situation. For some applications above, bias can be harmless while for others it have a lasting effect. Recommending which messages to display can be skewed towards one user or the other who may want instant popularity. Even recommending movies could be skewed towards movies that are seldom watched. These are fairly harmless biases and may not have lasting effect. But a rental website, consistently recommending a higher rent, or a home value estimation, consistently proposing a higher value can be detrimental in the long-term. I think these websites should provide a complete transparency of how they arrived at these recommendations. What do you think?