🧠AI Foundations
Your progress0%
0 of 49 lessons
Exercise10 min·Lesson 4 of 5

Try It Now — AI Claim Audit

🛠️
What you will need: A smartphone or computer, and access to either ChatGPT (chat.openai.com) or Google Gemini (gemini.google.com). Both have free tiers. You can also use WhatsApp if you have access to an AI-powered bot. This exercise takes about 10 minutes.

Your Task: Audit an AI Claim

You are going to find one real AI claim — from a news article, social media post, company website, or WhatsApp forward — and run it through the critical thinking framework you learned in this unit. Then you will use an AI tool to help you investigate it.

Step 1 — Find Your Claim

Look for an AI-related claim you have seen recently. Good places to look:

  • A WhatsApp forward about AI doing something impressive or scary
  • A LinkedIn post from a company announcing an "AI-powered" product
  • A news headline about AI in Kenya, Africa, or your industry of interest
  • A job advert that mentions AI skills

Write the claim down in one or two sentences. Example: "A Nairobi startup claims their AI can screen job applicants and rank the top 10% with 94% accuracy in under 30 seconds."

Step 2 — Run the Three Critical Questions

  1. What exactly does this AI claim to do? (Be as specific as possible. What is the input? What is the output?)
  2. How was it tested? (Is there any evidence of independent testing? Who did the testing? On what kind of data?)
  3. Who is harmed if it is wrong? (Think about which groups of people are most affected by errors — and whether those groups were represented in testing.)

Step 3 — Ask the AI Tool to Help You Investigate

Use Rafiki, your AI tutor, right here on this page — tap any ‘💬 Try with Rafiki’ button in this lesson, or the chat bubble at the bottom right. Replace the bracketed section with your actual claim:

💬
I found this AI claim: [paste your claim here]. Help me think critically about it. What questions should I ask about the training data? What are the most likely ways this claim could be misleading or overstated? Who might be harmed if the AI makes errors? Please be honest about what you do not know.

Read the response carefully. Notice: does the AI tool give you useful critical questions, or does it mostly validate the claim? Does it acknowledge its own limitations?

Step 4 — Check for Hype Signals

Go back to the five hype signals from Lesson 1 and mark which ones you found in your claim:

  • No numbers, or suspiciously round numbers
  • Lab results presented as real-world results
  • Use of "always" or "never"
  • No mention of failures or limitations
  • "Replaces" humans instead of "assists" humans

Step 5 — Write Your Verdict

In 3–5 sentences, write your verdict on the claim. Rate it on a simple scale:

  • Credible: Specific, independently tested, limitations acknowledged
  • Uncertain: Some evidence but important questions unanswered
  • Hype: Vague, untested, or cherry-picked evidence
🤔
Reflection prompt: Did the AI tool you used help you think more critically, or did it mostly agree with whatever you said? What does that tell you about how to use AI as a thinking partner? Share your verdict and reflection in the Hekima community — your analysis might help someone else who saw the same claim.