AI Bias in Practice

Learning Objectives Recognize the most common forms of AI bias in everyday tools Understand why AI bias happens and why it is so hard to eliminate Know what you can do about bias at the individual and organizational level Why AI Is Biased AI models learn from data. If the data reflects historical biases — and it almost always does — the model will reproduce and sometimes amplify those biases. This is not a bug that can be patched. It is a fundamental characteristic of how machine learning works. The training data for large language models includes billions of web pages, books, articles, and conversations that reflect centuries of human prejudice alongside human knowledge. The result: AI systems can produce output that is biased along lines of race, gender, age, nationality, religion, socioeconomic status, disability, and other dimensions — even when the people who built the system did not intend any bias. Real-World Examples Hiring and Recruitment AI resume screening tools have been shown to penalize candidates based on proxies for race, gender, and socioeconomic background — the name of the college, the zip code, the name of the applicant, even the hobbies listed. Amazon famously scrapped an AI hiring tool in 2018 after discovering it systematically downgraded female candidates, because it was trained on a decade of resumes from a predominantly male workforce. In 2026, similar issues persist in more subtle forms. AI tools that score "culture fit" or "communication skills" often encode the communication patterns of the majority…