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 do not have ethics. Their output is a result of the type of algorithm and the input provided that is provided. If the input is biased, there’s a good chance the output will be biased too. I have spent most of my career in banking and financial services. This is an industry that uses recommendation systems for loan approvals. If a recommendation system is trained using data that systemically discrimates based upon race or sex then the recommender system will (could) also discriminate unless the data scientist does something to prevent it. Let use the example of race and African American getting mortages. My first suggestion to ensure a fair and equitable process is to deny the system information about race - take that variable out of the equation. In Evan Estola youtube presentation, he said that Meetup segreates the demoagraphic information so that the recommendations are made with out the benefit of this information. I think something similar should be adopted by the banking industry.
My second solution to help prevent biased recommender systems is Diversity. I’m confident that a diverse team of data scientists is much more likely to produce alogrithms that aren’t biased compared to a group of five white men. That’s not to knock white men, but rathter a recognition of the power of diversity. To achieve this minorities and women need to pursue jobs in data science and machine learning and corporate america needs to make a concerted effort to hire these individual with the explicit goal of building diversity and unbiased algorithms.