AI ethics is the study of how artificial intelligence systems should be designed, tested, and used so they help people and avoid harm. AI systems can affect school recommendations, job screening, medical tools, search results, and social media feeds. Because these systems learn from data, they can repeat unfair patterns if humans do not check them carefully.
Ethical AI matters because technology should be accurate, fair, safe, and understandable.
Key Facts
- Machine learning finds patterns in data and uses them to make predictions or decisions.
- A basic model workflow is data input → training → prediction → evaluation → improvement.
- Bias can enter an AI system through unbalanced data, flawed labels, or unfair design choices.
- Accuracy = correct predictions / total predictions.
- False positive rate = false positives / actual negatives.
- Ethical AI should include fairness, transparency, privacy, safety, and human oversight.
Vocabulary
- Artificial Intelligence
- Artificial intelligence is computer technology that performs tasks that usually require human thinking, such as recognizing patterns, making predictions, or understanding language.
- Machine Learning
- Machine learning is a type of AI where a computer improves its performance by finding patterns in data instead of following only fixed rules.
- Bias
- Bias is a systematic unfairness in data, design, or results that causes an AI system to treat some groups less accurately or less fairly.
- Transparency
- Transparency means making an AI system's data, goals, limits, and decision process understandable to people who use or are affected by it.
- Privacy
- Privacy is the protection of personal information so it is collected, stored, shared, and used only in safe and appropriate ways.
Common Mistakes to Avoid
- Assuming high accuracy means the AI is fair. A model can be accurate overall but still perform poorly for a smaller group in the data.
- Using data without checking where it came from. Data can include missing groups, old patterns, or human labeling errors that affect the model's decisions.
- Treating AI decisions as automatically objective. AI systems reflect choices made by people, including what data to collect, what goal to optimize, and what errors are considered acceptable.
- Ignoring privacy because the data is useful. Personal data must still be protected, minimized, and used with permission when required.
Practice Questions
- 1 An AI model makes 200 predictions and 170 are correct. What is its accuracy as a decimal and as a percent?
- 2 A screening model is tested on 80 students in Group A and 80 students in Group B. It makes 8 mistakes for Group A and 20 mistakes for Group B. Find the error rate for each group and state which group the model performs worse on.
- 3 A school wants to use AI to recommend advanced classes. Explain two ethical checkpoints the school should use before trusting the recommendations.