How Bias and Ethics Affect Artificial Intelligence

What is bias?

Bias is a disproportionate favoring in preference when it comes to an object, person, or event. A bias can be made internally or externally and is a lack of objectivity when assessing something.?

Bias in Business

Situations, when bias can be an issue within a marketing or business environment, can come down to leadership decision-making. The problem recognition stage of ethical decision-making can be impacted by bias lowering the ethical environment of a company. In order to improve unethical business practices, leaders can implement strategies to help and through better training (LSE).?

Bias in Artificial Intelligence

Everyone possesses some form of bias including artificial intelligence. AI can have two forms of bias with the first being algorithmic bias where algorithms utilize biased data. The second kind of AI bias is a societal bias where our assumptions and norms as a society provide predetermined expectations (Lexalytics).?

An example of AI bias can be seen in mortgage applicants in the US. An investigation by The Markup looked into if people of color were getting their mortgage applications denied more than white people. The study showed that 80% of black applicants, 40% of Latino applicants, and 70% of Native American applicants are more likely rejected. This is very important as 45% of the US’s largest mortgage lenders are offering loans digitally (Forbes).?

Making Algorithms Bias-Free

It’s would be very difficult to produce algorithms that aren’t biased but one strategy for reducing bias would be feeding algorithms data sets that are perfectly corrected. Even with a perfect data set context would be another issue you would face as well. A recommendation that I would offer for dealing with bias is creating a bias detection alert for when an algorithm might face bias issues. This would prevent further ethical implications from snowballing (Brookings).?

Final Thoughts

Bias is natural and will always be around whether we like it or not. Removing bias as much as we can from ai algorithms is a difficult task but monitoring algorithms to notice bias would be a possible solution to minimizing the amount of bias we see in its output.

References