AI bias study: organizational pressure, not just data, drives unfair algorithms

The study by Micarah Malone-Gawu explores AI bias through the lived experiences of nine practitioners. Using semi-structured interviews and document analysis, the research identifies three primary sources of bias: historical inequities embedded in training data, exclusionary design assumptions from homogenous teams, and organizational cultures that reward speed over ethical reflection. Participants emphasized that bias is not a purely technical problem; it requires addressing systemic issues such as lack of diverse perspectives and weak governance frameworks. Many noted that ethical standards exist but are poorly enforced, leading to inconsistent responsible practice.

The paper concludes that achieving equitable AI demands a human-centered approach that integrates ethics from the earliest design stages. Recommendations include strengthening accountability structures, cultivating cognitive awareness among developers, and creating institutional environments that encourage critical reflection. The study contributes to ongoing responsible AI conversations by providing practitioner-grounded insights. For organizations, the message is clear: fairness cannot be patched in after deployment; it must be built in through deliberate policy, diverse teams, and a culture that values ethical rigor over speed.