The thermostat analogy
A thermostat doesn't need to be right. It needs to detect when the temperature is wrong and adjust. It measures, compares to a target, and acts. If you set it to 72 degrees and the room hits 75, it doesn't hold a committee meeting. It turns on the air conditioning.
Most government policies have no thermostat. They are designed, debated, enacted, and then left to run — sometimes for decades — with no built-in mechanism to detect whether they are achieving their intended outcomes. When they fail, the failure is usually discovered by journalists, advocates, or affected citizens long after the damage has compounded.
This is not a criticism of policymakers. It is a criticism of the policy design process. We build policies like blueprints (static, one-time) when we should build them like thermostats (dynamic, self-monitoring).
Six criteria for self-correction
We evaluate policies against six criteria, each weighted by its importance to the policy's ability to detect and respond to its own failures. The total score ranges from 0 to 100.
1. Feedback Mechanisms
20%Does the policy collect data on its own performance? Are there defined metrics, reporting requirements, and channels for frontline feedback? A policy without feedback mechanisms is flying blind.
2. Sunset Clauses
15%Does the policy expire? Sunset clauses force periodic reauthorization, creating natural review points. Without them, policies persist on inertia alone — even when the problem they were designed to solve has changed.
3. Measurable Outcomes
20%Can you tell if it's working? Specific, measurable targets with baselines and timelines. Vague goals like "improve public safety" score low. Concrete targets like "reduce response times by 15% within 24 months" score high.
4. Review Triggers
15%What forces review? Calendar-based reviews are adequate. Event-triggered reviews (when a threshold is crossed, when conditions change) are better. Best: both, with clearly defined authority to initiate review.
5. Adaptation Pathways
15%Can the policy change without full repeal? Policies that require legislative action to modify any parameter are brittle. Policies that delegate calibration authority to implementing agencies can adapt faster.
6. Stakeholder Input
15%Are affected parties heard? Formal comment periods, advisory boards, public reporting, and community feedback loops. The people closest to the problem often detect failures earliest.
Two policies, two scores
To illustrate how this framework works in practice, consider two policies scored against the six criteria:
| Criterion | Clean Energy Standard | Student Loan Forgiveness |
|---|---|---|
| Feedback Mechanisms | 18/20 | 6/20 |
| Sunset Clauses | 12/15 | 2/15 |
| Measurable Outcomes | 17/20 | 8/20 |
| Review Triggers | 13/15 | 4/15 |
| Adaptation Pathways | 12/15 | 5/15 |
| Stakeholder Input | 10/15 | 5/15 |
| Total | 82/100 | 30/100 |
Clean Energy Standard (~82/100): Built-in annual reporting requirements, percentage-based targets with measurable milestones, regular compliance reviews, authority for implementing agencies to adjust timelines based on grid conditions, and formal stakeholder comment processes. This is a policy with a thermostat.
Student Loan Forgiveness (~30/100):No built-in mechanism to measure whether the policy achieves its stated goal (reducing financial burden to expand economic participation). No sunset clause. No adaptation pathway short of new legislation. No formal review trigger. Limited stakeholder input after enactment. This is a policy without a thermostat — it either works or it doesn't, and there is no structural mechanism to tell which.
The score doesn't tell you whether a policy is good. It tells you whether a policy can learn. An 82 can still fail. A 30 can still succeed. But when they fail, the 82 will detect and correct. The 30 will persist unchanged until external pressure forces a reckoning.
Quick wins vs. structural changes
Not all improvements require new legislation. There is a meaningful category of improvements that agencies can implement administratively:
Administrative (no legislation needed)
- Add performance metrics to existing program reviews
- Establish regular stakeholder feedback sessions
- Define threshold-based review triggers within existing regulatory authority
- Create dashboards that track policy outcomes against stated goals
- Implement after-action reviews when programs miss targets
Structural (requires legislation)
- Add sunset clauses to new legislation by default
- Delegate calibration authority to implementing agencies
- Mandate outcome measurement as a condition of authorization
- Require adaptation pathways in policy design
- Create independent review bodies with trigger authority
The administrative improvements are where most agencies should start. They require no new authority, produce visible results within one budget cycle, and build the institutional muscle for the structural changes that follow.
Applying this to your work
Every policy your agency administers can be scored against this framework. The exercise itself is valuable — it forces a conversation about what "working" means for each policy, what data would tell you it's failing, and who has the authority to course-correct.
Most agencies that run this exercise discover that their highest-profile policies score lower than their routine ones. The policies that attract the most political attention are often the least self-correcting, because the design process optimized for passage, not for learning.