Bias in Data

Part of: Data

Data Is Not Always Neutral It is tempting to think numbers are objective, but data bias means a dataset systematically misrepresents reality. Biased data leads to biased conclusions, and when programs make decisions from that data, the bias is repeated at scale. The AP CSP exam expects you to recognize how bias enters data and why it matters. How Bias Enters Data - Sampling bias : the data does not represent the whole population. A survey taken only on a college campus will not reflect retirees. - Collection bias : how data is gathered skews it. A satisfaction poll sent only to people who finished a purchase ignores those who gave up. - Historical bias : data reflects past inequities, so a model trained on it can perpetuate them. - Exclusion bias : leaving out groups, such as a voice system trained only on certain accents. Why Bias Matters Biased data can cause real harm: unfair loan decisions, inaccurate medical predictions, or recommendation systems that exclude groups. Because programs scale, a small bias can affect millions of people quickly. Detecting Imbalance A first step is checking whether groups are represented fairly. Suppose we count responses per group and flag whether

Challenge: Detect Imbalanced Data