Module 05Free

Putting It Together — A Practical Diagnostic

A four-step method for working out why an AI answer went wrong, and fixing the real cause.

Teaching content

Module goal

Diagnose a disappointing AI answer against the four properties and apply a targeted fix instead of guessing.

Lesson 1The four pieces collide in real use

You now have four pieces of the puzzle: next-token prediction (how answers are built), knowledge (what the model learned, and its limits), working memory (what it can currently see) and steerability (how much your instructions control the outcome). In real use these don't operate one at a time — they collide. A long, complex question can push against working memory limits while also straying into knowledge the model doesn't have, while a vague instruction fails to steer it back.

Lesson 2The four-step diagnostic

Ask yourself	If yes, the likely cause
Is the answer confidently wrong on a specific fact?	A knowledge gap — the model is filling in a plausible guess. Verify against a real source, or ask the tool to search the web.
Has this conversation been very long, or did I paste a lot of text earlier?	A working memory limit. Restate the key facts, or start a fresh chat with only what matters.
Is the output generic, off-tone or off-target?	Weak steering. Add audience, tone, length, format, an example, and what to avoid.
Is the topic niche, local or very recent?	Thin training coverage. Treat the answer as a draft and confirm it independently.

Real-life example: an event planner gets a wedding theme suggestion that ignores her budget. Her targeted fix is to restate the budget explicitly (“The confirmed budget is GHS 3,500”) and give a specific instruction (“Suggest a theme that fits a 25-year-old bride who loves Ankara prints and bold colours, not neutral tones”). The next answer is accurate and genuinely useful — because she fixed the cause instead of repeating the same vague question.

The goal of this course

Not to make you distrust AI, but to give you a clear enough mental model that you know when to trust it, when to double-check it, and how to get dramatically better results either way.

Practical task

Try it yourself

Capstone: diagnose and fix

1.Think of a recent moment when an AI tool gave you an answer that felt off, confusing or wrong — or try a new task now.
2.Walk through the four-step diagnostic and identify, in your own words, which of the four properties was most likely the cause.
3.Try the task again with a fix based on your diagnosis.
4.Compare the two results and write one sentence on what actually changed.

Final quiz

Multiple choice • instant feedback

Question 1 of 5

When an AI gives a confidently wrong answer, what should you suspect first?

Mark your progress

That's the whole course. Head back to the overview to collect your AI Literacy Explorer badge.