Hold beliefs as estimates, not identity commitments

Treat each belief as a probability rather than a flag to plant.

Why it works

When a belief is fused with identity, new evidence against it reads as an attack on the self, triggering defensiveness rather than updating. Decoupling belief from identity — thinking "my current best estimate is X, at about 70% confidence" rather than "I am someone who believes X" — lowers the psychological cost of changing your mind. Galef calls this holding beliefs "loosely": with conviction proportional to evidence, not to social commitment.

How to do it

1. Express beliefs as probabilities rather than assertions: "I think this is probably true, maybe 75%."
2. Before a conversation where your view might be challenged, mentally separate your self-worth from the belief.
3. When evidence comes in, ask: "What does this do to my confidence level?" rather than "Am I winning or losing?"

Evidence

Belief-as-estimate framing underlies the calibration training in the Good Judgment Project, which produced measurably better forecasting accuracy in participants trained to think probabilistically. Mellers et al. (2014) isolate the effect experimentally: forecasters given probabilistic-reasoning training and belief-updating practice outperformed controls across a multi-year geopolitical tournament, evidence that holding beliefs as movable estimates is a trainable skill rather than a fixed trait. (observational)

Calibration research measures forecasting accuracy, not interpersonal outcomes; whether probabilistic belief framing improves relationships or conversations is a reasonable inference, not a directly studied effect.

Sources

• Tetlock & Gardner (2015), Superforecasting — calibration training and forecasting accuracy
• Mellers, B., Ungar, L., Baron, J., Ramos, J., Gurcay, B., Fincher, K., Scott, S. E., Moore, D., Atanasov, P., Swift, S. A., Murray, T., Stone, E., & Tetlock, P. E. (2014). Psychological strategies for winning a geopolitical forecasting tournament. Psychological Science, 25(5), 1106-1115.

Common mistake

Performing probabilistic language ("I might be wrong about this") while internally holding the belief as settled — the performance is detectable and the underlying rigidity remains unchanged.