Figure 7
From: A comparison of human and GPT-4 use of probabilistic phrases in a coordination game
Stability of GPT-4 estimates in the Medical Condition. Figure 2b (GPT-4 estimates of probability in the Medical Context) contains an evident outlier not certain whose estimate in the Investment Condition is 50%, in the Medical Context, 95%. We plot the GPT-4 estimates from the Medical Context in red with the outlier marked by a red disk. We repeat these GPT-4 estimates four times and plot them as well, in blue. The contours are slightly offset vertically so that they are visible. The outlier does not reoccur, and the four repetitions are in good agreement with each other and the original estimates except for the outlier. See text.
