Author: Data Science Ethics Podcast
Publisher: Spotify
Publication Year: 2020
Summary: The following podcast episode discusses how gender and racial gaps exist in the field of data science as it tends to be dominated by certain groups such as white men. While we understand that inclusivity and diversity are essential for ethical data science practices, constraints sometimes prevent teams from having a wide spectrum of backgrounds (e.g. small size of team). To promote an inclusive environment and model, we should consider who is working on the model, who gives feedback about the model, and who the model is tested with. Diversity does not always imply race or gender but personality types; therefore, StrengthsFinder can be used to assemble a team by various strengths to have a broader perspective about how people think and problem-solve. In addition to getting diverse perspectives within the team creating the algorithm, data scientists should also construct a diversified pool of end users to discuss the intended outcomes of the algorithm. This can be done through surveys or focus groups, but overall, this discussion is intended to gather insight into how nontechnical audiences may interpret and understand the impacts of the model being implemented. For example, end users can express how they may be adversely affected by the algorithm, thereby raising issues for the team to address before releasing the model. A method to receive unbiased feedback is providing limited information when initially exposing the model to allow users to form their own opinions and challenge assumptions.