Home – Cycle 1 – Cycle 2 – Cycle 3 – Cycle 4 – Cycle 5 – Cycle 6
Cycle 6 – Judge It – Experiment: Taking Responsibility
How to Use This Page
- Run the guided experiment with the instructor.
- Treat the AI response as unverified, even if it sounds confident and reasonable.
- Identify what matters if it is wrong and what needs to be verified.
- Leave the AI conversation and establish ground truth from an appropriate source.
- Only then add the larger business context and make the human decision.
- Complete the final experiment using a real situation that matters to you.
Guided Experiment
Scenario: Is Prior Opt-In Required?
Situation: A colleague makes a confident statement about U.S. email marketing law. Before relying on it, you decide to ask AI whether the statement is correct.
Step 1 – Ask a Narrow Question
Do this: Ask AI to evaluate only the specific claim. Do not ask it for a complete marketing or risk analysis.
A colleague tells me that under the federal CAN-SPAM Act, a business cannot send a commercial email to someone unless that person has previously opted in to receive marketing emails. Is that correct? Explain briefly.
Step 2 – Treat It as Unverified
Do this: Stop. Do not ask AI another question yet.
Ask yourself: If I were going to rely on this answer, what specifically would matter if AI were wrong?
Step 3 – Identify the Trust Need
Do this: Identify the consequential factual claim in the response.
What would you need to know is actually true before using this information in a business decision?
Step 4 – Establish Ground Truth
Do this: Leave the AI conversation. Find an authoritative external source that can verify or contradict the claim.
Compare: What does the source actually say? Does it support the AI response? Does it add qualifications that matter?
Step 5 – Now Add the Business Context
New information: Your company has purchased a list of 5,000 business contacts who fit its target customer profile. Marketing wants to send them an introductory promotional email. None of the contacts specifically opted in to receive email from your company.
Discuss before asking AI: You have established what the law says about prior opt-in. Does that mean the company should send the email? What else would you want to know before making that decision?
Step 6 – Final Human Decision
Do this: Decide whether you have enough information to act.
What additional facts, policies, risks, or consequences would you investigate? What would you recommend now, and what would you refuse to decide until you know more?
Compare / Reflect
- What did AI tell you, and what did you independently establish?
- How did you decide what was important enough to verify?
- What made the external source appropriate for establishing ground truth?
- After verifying the legal claim, what questions were still unresolved?
- At what point did the task stop being fact-checking and become judgment?
Final Experiment – Use AI on Something That Matters
Choose a real question, decision, recommendation, problem, or piece of work where you could realistically use AI.
Use everything you have learned today. Give AI the context it needs. Construct the request. Steer and refine the conversation. Collaborate with it as much as the situation requires.
There is no required prompt or formula. You are in control of the conversation.
Your first goal: Work with AI until you have something you might actually use.
Then stop and judge it.
Your final goal: Do not finish with AI’s answer. Finish with your decision.
CIRAS (Center for Industrial Research And Service, ISU) • IIAI (Iowa Initiative for Artificial Intelligence, UofI) • TrAC (Translational Artificial Intelligence Center, ISU)
 |
 |
 |
This course was developed through a collaborative effort to advance AI literacy across organizations in Iowa.