Figure 3
The two-panel grouped vertical bar chart is drawn on coordinate planes and compares the weight of advice under different advisory conditions. In both panels, the vertical axis is labeled “Weight of Advice” and ranges from 0 percent to 100 percent in increments of 20 percent. Panel (a), titled “G P T decisions compared to humans”, displays 4 bars, with 2 bars per category. The horizontal axis lists categories from left to right as “Algorithm Advice” and “Human Advice”. A legend identifies two groups as “L L M s” and “Humans”. For “Algorithm Advice”, L L M s have a weight of 48 percent and Humans have a weight of 47 percent. For “Human Advice”, L L M s have a weight of 80 percent, and Humans have a weight of 32 percent. Panel (b), titled “G P T W O A across models and temperature”, displays 16 bars, with 4 bars per category. The horizontal axis lists categories from left to right as “Algorithm Advice 20 percent”, “Human Advice 20 percent”, “Algorithm Advice 50 percent”, and “Human Advice 50 percent”. A legend identifies four groups as “G P T-3.5, Temp equals 0”, “G P T-3.5, Temp equals 1”, “G P T-4, Temp equals 0”, and “G P T-4, Temp equals 1”. For “Algorithm Advice 20 percent”, the weights are: G P T-3.5 Temp 0: 45 percent; G P T-3.5 Temp 1: 33 percent; G P T-4 Temp 0: 52 percent; G P T-4 Temp 1: 50 percent. For “Human Advice 20 percent”, the weights are: G P T-3.5 Temp 0: 90 percent; G P T-3.5 Temp 1: 83 percent; G P T-4 Temp 0: 80 percent; G P T-4 Temp 1: 92 percent. For “Algorithm Advice 50 percent”, the weights are: G P T-3.5 Temp 0: 60 percent; G P T-3.5 Temp 1: 65 percent; G P T-4 Temp 0: 40 percent; G P T-4 Temp 1: 44 percent. For “Human Advice 50 percent”, the weights are: G P T-3.5 Temp 0: 95 percent; G P T-3.5 Temp 1: 92 percent; G P T-4 Temp 0: 49 percent; G P T-4 Temp 1: 53 percent. Note: All numerical values are approximated.Panel (a) compares mean WOA for human vs. algorithmic advice in GPT and humans from Logg et al. (2019). Panel (b) plots WOA by the four factors. Source: Authors’ own work