Fair or Flawed? A Review Paper on Bias in Artificial Intelligence

Artificial Intelligence, or AI, is now used in many parts of daily life. It helps recommend videos, check passport photos, filter job applications, detect fraud, translate languages, and even support decisions in areas like healthcare, policing, education, and banking. AI is often seen as smart, fast, and neutral. However, AI is not always fair. Because AI systems are created by humans and trained on human data, they can learn the same unfair patterns that already exist in society. This review paper studies the problem of bias in AI. Algorithmic bias happens when an AI system produces unfair results for certain people or groups. This may happen because the training data contains social bias, because some groups are underrepresented in the data, because the system measures the wrong thing, because feedback loops make old patterns stronger, or because people purposely manipulate the data. These problems can cause serious harm, especially when AI is used in sensitive areas like hiring, housing, loans, policing, healthcare, and education. The paper explains five major causes of AI bias in simple language. It also discusses examples such as gender stereotypes in search results, facial recognition problems, biased predictive policing, AI essay grading, hiring systems, and manipulated chatbots. The paper further reviews why AI bias is difficult to fix. Some AI systems are hard to understand because they work like “black boxes.” Also, collecting more data about protected groups may help test fairness, but it can also raise privacy concerns. The paper concludes that AI should not be trusted blindly. Humans must test, monitor, and question AI systems before using them in important decisions. Fair AI requires better data, transparency, human oversight, ethical design, and strong rules. AI can be useful, but only if society makes sure it is used carefully and responsibly.