Coding Bias and What It Can Do to Our Future

Ideally, the way programmers train the algorithms should make our lives significantly easier and more manageable, saving a lot of waiting and processing time across all systems and functions.

However, because machine learning mechanisms, especially their early forms, were designed by a homogeneous group of individuals with their own biases, the algorithms included flaws that would compute information at someone's disadvantage and undeservedly so.

Some of such examples were well demonstrated in the Coded Bias (2020) documentary:

Scenario 1, where citizens of Britain were monitored and screened without an explicit announcement. It was not mentioned anywhere within the jurisdiction and was done in a rather sneaky way. The algorithms with which the human profile was analyzed were severely biased towards individuals of a certain race and demographics. And the mere fact that such surveillance could take place in a civilized democratic society is alarming. What if it signifies the beginning of a totalitarian government form where human rights are systematically violated, and where prejudiced judgments are made swiftly, without a chance to appeal and rehabilitate one's good name and reputation?

Scenario 2: Candidates who applied to Amazon (tech?) jobs were filtered by gender, and the ones who got hired were predominantly male because of the way the AI tool was trained to discern information. For example, the Resumes where the word woman was present in any form or context were weeded out.  This also meant, however, that candidates with more gender neutral Resumes, where first names could belong to either gender, and where organizations' names did not mention the word woman, had the chance to be invited to an interview.

Both of these scenarios demonstrate dangerous trends, and we do need systems in place that will overcome such biases and ensure that they won't penetrate our social infrastructure so deeply that it would be nearly impossible to eradicate in the later stages.

Coding should replicate the real world with its sentiments and help meet its vital needs, and not copycat its existing problems into new algorithms.