Do (not) enter? Objective visual features of indoor scenes predict approach-avoidance responses and core affect

Environmental psychology theories on perception and preference have come a long way to predict emotional and behavioral responses to environmental scenes, yet the gap between objective scene characteristics and human experience remains. Most existing studies relied on subjective qualifications of perceptual attributes, or examined the influence of few objective visual features at a time, capturing only a small part of the complexity of real world scenes. The novelty of this study lies in drawing from visual perception and advancements in computer vision to objectively quantify a large set of visual features from low-, mid-, and high-level vision on 360 photographs depicting everyday indoor environments and examine both the individual and combined effects of these visual levels on affective and approach-avoidance responses. These features were used in bootstrapped stepwise regressions to predict subjective responses of approach-avoidance, valence, and tense and energetic arousal, collected from 60 participants in a controlled environment. Results show that low-, mid-, and high-level visual features accounted for a substantial amount of variance in all outcome measures, jointly as well as per separate level. Subsequent variance partitioning analyses showed that explanatory power was shared between visual features from different levels with a relationship that depended on the target variable. Certain visual features, such as the number of edges, degree of enclosure, and presence of plants confirmed relationships proposed by key theories of environmental preference, while others, such as mean brightness, emerge as important predictors despite not being discussed by the theories. This work demonstrates the capabilities of objectively quantifying features from all levels of visual perception in environmental preference research, and advances our understanding of the relationship between visual features and emotional responses to indoor environments, highlighting the need for more detailed theories in this domain.