What has to be true?

Every plan has a hidden theory inside it. “If we write an op-ed, the board will change the policy” assumes three things. The board reads op-eds. The board cares about public opinion on this issue. The board has the authority to change the policy. If any one is false, the plan fails, however well you write.

A theory of change (show definition) makes that hidden theory visible. The evaluator Carol Weiss made the term popular in 1995. She described it as the set of assumptions that link a program’s small steps to its long-term goal.

Write yours in one line: If we do this, then that will happen, and here is why.

Here is a made-up example.

• Goal: the district adopts a policy on AI-made fake images of students.
• Plan: collect 300 student signatures, then speak at public comment.
• Theory: if 300 students sign, the board will put the policy on its agenda.
• Assumptions: the board responds to petitions. 300 is enough to matter. The board, not the state, controls this policy.
• Weakest assumption: the first one. Nobody has checked.
• Cheapest test: email one board member and ask how items get on the agenda. Cost: one email.

Now a real one. OpenAI named its agents in the Hugging Face hack on July 21, 2026. Two days later, two members of Congress, one from each party, introduced the AI Kill Switch Act. Their bet: a public scare opens a window for a law. By October 2026 the bill had 11 cosponsors, and it had not left a subcommittee. What else has to be true for it to pass?

The deck’s “In or Out?” debate shows four strategies. Inside: join the committee. Outside: push with petitions and press. Build: make the better tool. Teach: change one class at a time. The right one depends on which assumptions hold where you live.