What Is Bias?

Imagine a coin that lands heads 70 percent of the time. You would say that coin is biased — it does not give both sides an equal chance. Now imagine a hiring system that rates identical resumes differently depending on the name at the top. The system is biased in the same way: it tilts unfairly toward or against people based on something that should not affect the decision. In this lesson, you will learn exactly what bias means, why it matters, and how to spot it.

Bias in Everyday Life

Bias is not a strange, rare thing. It shows up everywhere humans make judgments — in which students teachers call on most often, in which neighborhoods get more police patrols, in which job candidates get callbacks. Bias means a systematic tilt: a pattern that consistently favors or disfavors a group in a way that is not justified by anything fair or relevant. The key word is systematic. A single mistake is not bias. Bias is a repeating, predictable pattern. If a teacher consistently calls on students in the front row and rarely on students in the back, that is a biased pattern — even if no individual decision felt unfair.

Bias can come from many sources. It can come from explicit prejudice — someone who consciously dislikes a group. But most bias in complex systems is subtler. It comes from assumptions baked into rules, from historical patterns treated as natural, or from data that reflects an unequal world. This subtler kind of bias is especially important for understanding AI.

Relevant vs. Irrelevant Attributes

When we say a tilt is unfair, we mean it is based on an attribute that should not affect the decision. The key question to ask is: is this characteristic relevant to what we are actually trying to measure or decide? A bank deciding whether to approve a loan should look at an applicant's income, debt, and credit history — those are relevant. It should not look at the applicant's race or gender — those are irrelevant to creditworthiness. If the bank's system gives worse scores to applicants of a certain race, even when their finances are identical to someone else's, that is bias. This distinction — relevant vs. irrelevant attribute — is the core of almost every fairness argument you will ever encounter.

Why Bias in AI Is a Serious Problem

AI systems are being used for high-stakes decisions: whether you get a job interview, whether you are approved for a loan, how likely a judge thinks you are to re-offend, whether a medical scan flags a disease in your body. When an AI system has bias, those biased outcomes scale up enormously — a biased algorithm applied to millions of people produces millions of biased outcomes. This is different from a single biased person making a single bad call. A biased human hiring manager might affect dozens of candidates a year. A biased AI hiring filter might affect millions of applications in the same period.