What I am most proud of in my work with the Truth Review is the “Evidence Mapping” step I devised in my fact-checking tool. It makes the user practice pinpointing counter-evidence. But can data itself be biased?
Algorithmic bias is always a hot potato when debating on issues related to hiring or policing. Where I spot it most often is sports and education. Historical data is what makes the backbone for scouting models. If the numbers say “Pitchers under 5’10” are prone to injuries and therefore rarely succeed,” there will never be another pitcher like Tim Lincecum or Marcus Stroman who makes it to the big leagues.
I believe this is “Moneyball” at its worst. We see a feedback loop. Suppose a model concludes that “Students from this region struggle academically in college.” Then the university stops would be less interested in looking for students in that region. And following this, less students are admitted from the region and the data would be “confirmed.”
It’s something I like to call “Zombie Data” in policy analysis. It’s dead but eats the living potential of our world. I see many educational institutions treating students this way, seeing them as mere datasets even if the data may be dead. Students basically seem to be viewed as prospective WAR (Wins Above Replacement).
I believe the solution to this is a “Human-in-the-Loop” policy. Similar to how MLB teams use data as one source of information for scouts to make their decision, educational and social policies should focus on integrating AI and tech into decision-making as a resource instead of the sole decision-maker. We need to be the ones revealing the hidden potential in underprivileged communities, not just choose the low-risk path. This is why I am hoping to continue to develop my idea on a “Bias Checker” through the Truth Review. This is a tool that helps the user check what evidence they may have excluded and how it could have altered the content inside the source.
We are in a world ruled by algorithms. It’s not something as bad as it sounds. But it’s going to be bad if we continue to go in the current direction. The secret to fixing this? Look at outliers. Look at data points way off the chart. Recognize their value.