Challenge your theory of change assumptions

The problem

Your theory of change looks logical on paper but you're too close to see the assumptions you're making. You say 'skills training leads to employment' but what about barriers like childcare, transport, discrimination? Your TOC assumes linear progression but real life is messy. Assumptions are invisible to you because they're baked into your thinking. You need someone to ask the awkward questions you're not asking yourself.

The solution

Use Claude or ChatGPT to act as a critical evaluator of your theory of change. Paste your TOC and ask: what assumptions am I making? Where's the logic weak? What barriers am I not accounting for? What evidence would I need to prove this works? What could go wrong that I haven't considered? The AI spots gaps in your logic, questions your causal claims, identifies missing steps, and flags where you're being optimistic without evidence.

What you get

A list of critical questions and gaps in your TOC: 'You assume skills → employment but don't account for: employer discrimination, lack of local jobs, competing with more experienced candidates. Your TOC shows linear progression but doesn't address: people dropping out, setbacks, non-linear journeys. Missing evidence for: claim that confidence building leads to job-seeking (what's the mechanism?). Unexamined assumption: employers will hire your beneficiaries.'

Before you start

• A draft theory of change (even a rough one)
• Willingness to have your logic challenged
• Openness to identifying gaps and weaknesses
• A Claude or ChatGPT account

When to use this

• You're developing a new theory of change and want to stress-test it
• Your existing TOC hasn't been challenged critically in years
• You're applying for major funding and need to demonstrate robust impact logic
• You're too close to your work to see your own assumptions

When not to use this

• You haven't thought through the basics yet - do that first before seeking critique
• You're not willing to genuinely reconsider your approach
• Your TOC is already being rigorously evaluated by external experts
• You're seeking validation rather than challenge - AI will point out weaknesses

Steps

1. 1

   Write your theory of change first

   Don't use AI to write your TOC. You know your work, your beneficiaries, what you're trying to achieve. Write it yourself: inputs → activities → outputs → outcomes → impact. Include your assumptions. Get it to a point where you think it makes sense. That's when critique is most valuable.

2. 2

   Ask for assumption challenges

   Paste your TOC into Claude or ChatGPT. Ask: 'What assumptions am I making in this theory of change? What am I treating as given that might not be true? Where am I being optimistic about how change happens? Be specific about which assumptions are most questionable.' The AI will spot things you've assumed away.

3. 3

   Question the causal logic

   Ask: 'Where is the causal logic weak in this TOC? Where do I claim A leads to B without explaining the mechanism? What steps am I skipping? What barriers might prevent the change pathway working as described?' Your TOC might say 'training → skills → employment' but what about the steps in between?

4. 4

   Identify missing perspectives

   Ask: 'Whose perspectives are missing? Have I considered: beneficiary barriers (systemic, practical, personal), external factors (economy, policy, discrimination), unintended consequences, why this might fail for some people? Which stakeholders' views aren't reflected?' You can't see all angles alone.

5. 5

   Challenge attribution claims

   Ask: 'Where am I claiming credit for changes that might happen anyway or be caused by other factors? How would I distinguish between: correlation vs causation, my impact vs external changes, short-term effects vs lasting change? What evidence would prove my contribution?' This is about honest impact assessment.

6. 6

   Question success assumptions

   Ask: 'What am I assuming about why this will succeed? Do I assume: people will engage as planned, the service will be delivered as designed, external conditions will remain stable, resources will be sufficient? Where might reality diverge from my model?' Plans meet reality and reality usually wins.

7. 7

   Identify evidence gaps

   Ask: 'For each causal claim in my TOC, what evidence do I have? Where am I making claims based on hope vs data? What would I need to measure to prove this works? What research exists on this intervention approach?' Evidence-based TOCs are stronger than aspiration-based ones.

8. 8

   Strengthen your TOC

   Now revise: acknowledge assumptions explicitly, add missing steps in the causal chain, address barriers you hadn't considered, identify where you need evidence, make attribution claims more honest. Your TOC should be more robust and realistic. You're not changing your mission based on AI, you're strengthening your logic.

Tools

Resources

Part of this pathway