Similarity judgements: the comparison of normative predictions and subjective evaluations - A study of the ratio model of similarity in social context

Abstract

Introduction: This study examines the consistency between subjective similarity evaluations and the theoretical predictions derived from Tversky's ratio model of similarity, alongside the impact of additional positive and negative features on perceived similarity to ideal and bad politicians.

Methods: Using a sample of 120 participants, we assessed the similarity of eight candidate profiles to an ideal and bad politician, varying in positive and negative features. Participants' subjective evaluations were compared with theoretical predictions derived from Tversky's ratio model. The analysis focused on how candidate and referent valence influenced observed versus theoretical similarity.

Results: Subjective similarity judgments deviated systematically from theoretical predictions, especially for positively featured candidates, indicating a negativity effect. Additional positive features decreased the perceived similarity of favorable candidates to an ideal politician, while additional negative features did not significantly affect similarity judgments of unfavorable candidates.

Discussion: Our findings underscore a significant disparity between subjective and objective similarity judgments, notably for favorable candidates. While the ratio model performs well for unfavorable candidates, its applicability diminishes for favorable ones, emphasizing the role of feature valence in decision-making. Further research on feature valence is crucial for a comprehensive understanding across contexts.

Keywords: objective similarity; positive-negative asymmetry; ratio model; similarity judgements; subjective similarity.

FIGURE 1

The figure illustrates the objective (theoretical) and subjective (observed) similarity to both an ideal and bad politician for candidate profiles differing in the number of positive and negative features attributed to them (x-axis). Dashed lines depict the theoretical predictions, while solid lines represent empirical findings. Lighter lines indicate similarity to an ideal politician, whereas darker lines signify similarity to a bad politician.

FIGURE 2

The differences between observed and predicted similarity measures for similarity to an ideal and bad politician for favourable and unfavourable candidates. Positive values indicate that the observed similarity values are higher than the predicted values.