Figure 5.

Correlations between proposed heuristics for predicting information gain and the actual information gain. (a) Correlations for experiment 2. Expected utility (as calculated using optimal game design) has the highest correlation, and number of unique chests is the only heuristic with a significant correlation to actual information gain. (b) Correlations for experiment 3. The correlations to information gain for the value of the trap chest and for expected utility are of similar magnitude, but only expected utility is consistently predictive across the two experiments.