The Fewer Reasons, the More You Like It! How Decision-Making Heuristics of Image Quality Estimation Exploit the Content of Subjective Experience

Abstract

Imaging science has approached subjective image quality (IQ) as a perceptual phenomenon, with an emphasis on thresholds of defects. The paradigmatic design of subjective IQ estimation, the two-alternative forced-choice (2AFC) method, however, requires viewers to make decisions. We investigated decision strategies in three experiments both by asking the research participants to give reasons for their decisions and by examining the decision times. We found that typical for larger quality differences is a smaller set of subjective attributes, resulting from convergent attention toward the most salient attribute, leading to faster decisions and better accuracy. Smaller differences are characterized by divergent attention toward different attributes and an emphasis on preferential attributes instead of defects. In larger differences, attributes have sigmoidal relationships between their visibility and their occurrence in explanations. For other attributes, this relationship is more random. We also examined decision times in different attribute configurations to clarify the heuristics of IQ estimation, and we distinguished a top-down-oriented Take-the-Best heuristic and a bottom-up visual salience-based heuristic. In all experiments, heuristic one-reason decision-making endured as a prevailing strategy independent of quality difference or task.