As more systems and sectors are driven by predictive analytics, there is increasing awareness of the possibility and pitfalls of algorithmic discrimination. In what ways do you think Recommender Systems reinforce human bias? Reflecting on the techniques we have covered, do you think recommender systems reinforce or help to prevent unethical targeting or customer segmentation? Please provide one or more examples to support your arguments.
Recommender systems are everywhere and they cannot be avoided. Research has shown the recommender systems tend to nudge users to the extremes via getting funneled into feedback loops. For example, this consistently happens on my personal YouTube feed where if I watch one certain video, there will be several recommendations on similar videos. The course work for this class had assignment lectures to several YouTube videos and currently YouTube is recommending me a slew of Data Science videos related to recommender systems, machine learning, etc. With each view of a video more recommendations come up. This is just one example of many that recommender systems keep reinforcing user bias.
While is what the system is designed to do, it can become dangerous in some cases. For example, take the scenario of a user that is feeling suicidal. What content would the system recommend in this case. Other such scenarios are political views, terrorism, religious extremism, misinformation, and other controversial subjects.
This is an evolving field but there are several things that can be done to remove bias, such as manual moderation, using AI to remove remove bias. One example of is Project Redirect by Google Jigsaw, that determines if a user is casually browsing extreme violent/terrorist content or has more motivations behind viewing this type of content. In these cases, the system recommends non-violent videos aimed to de-radicalize them.
How to handle such scenarios can be and are ethical landmines, but there is definitely a need to solve these pressing issues brought about by recommendation systems.
Reference: https://www.wired.com/story/creating-ethical-recommendation-engines/