Understanding people’s preferences toward the fair distribution of limited resources, such as vaccines for preventing infectious diseases during a pandemic, is of fundamental importance. While there are extant distributive principles that can be used to explain people’s distributive preferences, predicting actual human distributive judgments is often challenging because people’s judgments are influenced by the rich semantic information of judgment targets, which often cannot be captured by the existing distributive principles. The current paper focuses on the prediction and understanding of vaccine allocation judgments during a pandemic, investigating people’s decisions on prioritizing certain occupations for vaccine distribution. To quantify rich semantic meanings associated with more than 130 occupations, we used word embeddings obtained from large-scale text data. The results showed that using word embeddings as regression predictors enabled reasonable predictions over people’s vaccination priority judgments for various occupations. It was also indicated that the interpretation method introduced by previous studies, which allows for exploring semantic dimensions involved in people’s judgments, did not substantially depend on the choice of regularization parameters of the regression analysis. Taken together, the current paper sheds light on the effectiveness of prediction-oriented judgment research while taking a look at its interpretability.

特集 : 人間科学専攻特集号
寄稿論文