Hold models loosely

Believe your model enough to act on it, but not so hard that evidence cannot update it.

Why it works

Psychological investment in a model ("I am a person who believes X") activates identity protection when the model is challenged — not just epistemic resistance, but ego defense. Framing models as provisional tools rather than identity claims reduces the cost of updating, making revision feel like intellectual progress rather than defeat.

How to do it

1. Use language that signals provisionality: "my current model says…" or "I think X, but I’m not certain."
2. Distinguish beliefs by confidence level and revise lower-confidence ones first when evidence arrives.
3. Celebrate successful updates — "I updated my map" — rather than treating them as failures.
4. Separate your self-worth from the accuracy of your model so updates do not feel like attacks.

Evidence

Research on belief perseverance and identity-protective cognition shows that people process threatening information less carefully, arguing against rather than evaluating it. Framing beliefs as provisional can reduce this response. (observational)

Telling people to hold beliefs loosely is easier than doing it; the identity-protection mechanism operates partly below awareness and is not easily switched off by instruction alone.

Sources

• Kahan et al. (2011), identity-protective cognition in science communication, Journal of Risk Research

Common mistake

Claiming to hold the model loosely while defending every implication of it — provisional language without provisional behavior.