Down the
Rabbit Hole
Dozens of papers, and they disagree. Then I stopped reading conclusions and started reading methods.
The puzzle
Traditional game theory says no one should trust strangers with their money. The rational move is to keep it. Yet in Berg, Dickhaut and McCabe's (1995) Trust Game, people invest. Consistently. Across cultures.
Social preferences
Altruism, reciprocity, fairness norms. We trust because we care about others.
Risk attitudes
Different perceptions of uncertainty. Maybe trust is just how we handle social risk.
The confusion
The literature is full of contradictory findings
Then I saw it
I started tracking what each study actually measured. Two methodological problems kept appearing.
1Different stakes and probabilities
Stakes and probability ranges varied wildly: small bets in one study, large ones in the next, extreme probabilities here, middling ones there.
Insight. Prospect Theory has different parameters for different contexts. Maybe they were testing different parts of the curve?
2Mismatched uncertainty
Studies were comparing apples to oranges. In risk, you know the dice is fair (known probabilities). In trust, you have no idea (ambiguity).
Insight. These are fundamentally different types of uncertainty. We need to control for this.
Can Prospect Theory explain trust, once you control for stakes, probabilities and uncertainty?
This became the thesis
The framework: Prospect Theory
What needed to happen
- 1Matched probabilities. Trust and Risk games with identical outcome probabilities.
- 2Controlled uncertainty. Match the type of uncertainty: known probabilities for both.
- 3Optimal design. Use Fisher information to select highly informative trials.
- 4Model comparison. Compare candidate models to see what really differs.