A Simulating Noisy Preferences

We revisit survey results from Section 6, and explore the impact of label noise on our proposed method. After some familiarization, we asked the participants to determine if a given scenario was fair. PCA is most sensitive to noise. In contrast, average regret and payments are comparable irrespective of input label noise. We compare the real distribution to the probit model in Figure 6 using a Q-Q plot.