Trustworthiness judgments without the halo effect: A data-driven computational modeling approach

Area of Science:

• Psychology
• Computer Vision
• Social Neuroscience

Background:

• Perceived trustworthiness in faces is often linked to attractiveness.
• The specific visual cues differentiating trustworthiness from attractiveness remain unclear.

Purpose of the Study:

• To identify visual cues for perceived trustworthiness independent of attractiveness.
• To investigate the role of approachability and facial expression in trustworthiness judgments.

Main Methods:

• Development of data-driven models to manipulate perceived trustworthiness in faces.
• Experimental designs (subtraction and orthogonal models) to control for attractiveness.
• Human judgments and machine learning algorithms to assess facial perceptions.

Main Results:

• Faces manipulated for trustworthiness were perceived as more trustworthy, but not more attractive.
• These manipulated faces were also rated as more approachable and having more positive expressions.
• Machine learning algorithms corroborated the increased perception of approachability and positive affect.

Conclusions:

• Visual cues for trustworthiness and attractiveness can be dissociated.
• Apparent approachability and facial emotion are key drivers of trustworthiness judgments.
• These factors may also influence general evaluations of facial valence.