Introduction to AI
Senior 1 · ICT
Introduction to AI
Lesson 4 of 5
22 min
A knife can prepare a meal or hurt someone; what matters is how it is used. AI is the same. It can help a doctor spot an illness earlier, and it can also treat people unfairly, spread lies or expose private information. AI ethics is the study of how AI should be built and used so that it is fair, safe and respects people. It matters to you because you will use AI tools at school, and you will live in a society shaped by them.
AI should support human decisions, not replace human responsibility. A person must check what the AI suggests, be able to overrule it, and answer for the final decision. This is often called keeping a human in the loop.
Worked example
An AI system reads a chest X-ray and reports "90% likely: pneumonia". The doctor examines the patient, compares the X-ray with the symptoms and decides on the treatment. If the AI was wrong, the doctor can catch the mistake. The doctor, not the software, is responsible for the decision.
Bias in AI means that a system gives unfair results for some groups of people. It usually comes from the training data: an AI learns from examples, so if the examples are unbalanced, so is the AI.
Worked example
In 2018, the researchers Joy Buolamwini and Timnit Gebru tested commercial systems that guess a person's gender from a photo. For lighter-skinned men the error rate was below 1%; for darker-skinned women it was as high as 34.7%. The systems had learned mostly from photos of lighter-skinned people. Nothing about darker skin makes faces harder to recognise; the data was simply unbalanced.
Another example: a speech system trained almost only on English voices works badly for Kinyarwanda speakers. The fix is more varied and balanced data, and testing the system on every group of people who will use it.
AI systems are hungry for data, and much of it is personal data: names, photos, voices, locations, health records, messages. Rwanda's 2021 law on the protection of personal data and privacy sets rules for how organisations may collect and use such data, and Rwanda's National AI Policy (2023) requires AI systems to follow that law.
For you, it means: do not type your full name, address, phone number, passwords or other people's private information into an AI chatbot. What you type may be stored and used to train future systems.
Keza asks a chatbot to write her essay on the history of her district. She then:
She has used AI responsibly: honest, checked and still her own work.
Tip
Use AI the SAFE way: Stay in charge (a human decides), Avoid sharing private data, Fact-check and watch for unfairness, Explain honestly when you used AI.
A bank uses AI to decide who gets a loan. Who should be responsible for the final decision, and why?
Solution: A bank employee. AI can make mistakes or be biased, so a person must check the result, be able to overrule it and answer for the decision.
A job-application AI was trained on the CVs of people hired in the past, most of whom were men. It now scores women's CVs lower. Explain why, and suggest a fix.
Solution: The AI learned the pattern in its unbalanced training data (past hiring favoured men), so it copies that unfairness: this is bias. Fix: train it on balanced data, test it on both groups, and have people review its decisions.
Name three kinds of information you should never type into an AI chatbot.
Solution: Any three of: passwords or PINs; your home address; your phone number; your full name with your school; other people's private information; health details.
Ishimwe uses an AI app to make a video in which a classmate appears to say something embarrassing. Which principles does he break?
Solution: Do not harm (it is a deepfake used to bully), privacy (he used his classmate's image without consent) and honesty. It may also break Rwanda's laws on cyber crimes.
Summary
Sources and credits
Facts about Rwanda's National AI Policy are from public announcements of its approval (facts only; no text reused). The SAFE checklist, the worked examples and the practice items were written by Ganzaa.
This lesson adapts CC BY-SA material and is shared under CC BY-SA 4.0.
