← THE JOURNEY
PHASE 02EXPERIMENTAL DESIGN

The
Hypothesis

Same bet. Different source. If you see the exact same probabilities in both games, any difference in behaviour must come from who is on the other side, not from the maths.

The innovation: an inverted Trust Game

The Trust Game of Berg et al. (1995) has a problem: you don't know the probabilities. You just know someone could betray you. That's second-order risk, uncertainty about uncertainty.

Our modification gives you explicit probabilities. First-order risk. You see exactly what could happen and how likely it is. The only question: is it a person behind that probability, or a lottery?

SOCIAL CONDITIONTRUST
  1. 01Invest with Player KV, a previous participant
  2. 02The amount is multiplied
  3. 03You see the probability that KV returns money, based on their past behaviour
NON-SOCIAL CONDITIONRISK
  1. 01Invest in a lottery
  2. 02The amount is multiplied
  3. 03You see the same probability, determined by the lottery
THE CRITICAL POINT

Identical probabilities, stakes and structure

The same 200 trials are played in both conditions. Only the framing differs: a previous participant's behaviour, or a lottery. If Prospect Theory parameters differ, it is because social context changes how you process risk.

STRUCTURE
Trials400, in 8 mini-blocks of 50
Ordersocial and non‑social alternate
Choiceinvest, or not
RANGES
Investment5 to 10 tokens in the final trials
Probability5% to 95%, in 5% steps
Return0 to 39 tokens, ×4

What we're testing: twenty candidate models

Cumulative Prospect Theory has two families of parameters. Value weighting: α (outcome sensitivity) and λ (loss aversion). Probability weighting: γ+ for gains and γ− for losses.

We fit twenty models per participant: sixteen nested Prospect Theory variants, one for each subset of {α, λ, γ+, γ−} allowed to differ between the social and the non-social condition, and four alternatives that are not Prospect Theory at all (expected utility, and three simple heuristics based on the probability, the gain or the loss alone)

H1
Model comparison
The best-fitting model is the full Prospect Theory model, with all four parameters estimated separately for the social and the non-social condition
H2
Social changes value weighting
At least one of α (outcome sensitivity) and λ (loss aversion) differs between conditions
H3
Social changes probability weighting
At least one of γ+ (gains) and γ− (losses) differs between conditions
H4 SECONDARY
Same story with reaction times
The differences hold when the parameters are estimated with a drift-diffusion model that fits choices and reaction times together

The hypotheses are non-directional. H2 and H3 are only tested for the parameters that the winning model lets differ. The whole plan is preregistered on OSF.

NEXT Picking the most informative of 17,955 possible trials
BUILDING THE MACHINE →
← PHASE 01Down the Rabbit Hole