Beyond the Algorithm: Professional Experiences and Perceptions of AI Bias

Bias in AI stems from human choices, not just data: The study found that algorithmic bias emerges from "historical inequities, exclusionary design assumptions, and organizational pressures"—meaning bias is built in through decisions made by people during development, not simply inherited from training data alone.

Technical fixes are insufficient without structural change: Participants emphasized that "technical corrections alone cannot ensure fairness," indicating that companies cannot solve bias problems through code alone; they need organizational accountability systems, diverse teams, and ethical oversight throughout the development process.

Organizational culture and incentives matter as much as individual ethics: The research revealed "limited enforcement of ethical standards and organizational cultures that inconsistently support responsible practice," suggesting that even well-intentioned AI practitioners face pressure to prioritize speed and efficiency over ethical reflection.

Diverse participation and early ethical planning are essential: The study concludes that "equitable AI requires structural accountability, diverse participation, and sustained cognitive awareness during the development lifecycle"—meaning companies must include diverse perspectives from the start and maintain ethical focus throughout, not add it as an afterthought.

Practitioners lack adequate support for responsible AI development: The nine interviewed professionals described working within systems that don't consistently support responsible practices, highlighting a gap between individual practitioners' awareness of bias risks and their organizations' commitment to addressing them.