Modelling perceptions of criminality and remorse from faces using a data-driven computational approach

Area of Science:

• Psychology
• Law and legal studies
• Computer science

Background:

• Perceptions of criminality and remorse are crucial in legal decision-making, affecting lineup selections, verdicts, and sentencing.
• Previous research has not fully elucidated the specific facial information driving these perceptions.

Purpose of the Study:

• To identify and visualize the facial information that contributes to perceptions of criminality and remorse.
• To develop data-driven computational models of criminal and remorseful facial appearances.
• To explore correlations between these new models and existing facial perception models (dominance, threat, trustworthiness, etc.).

Main Methods:

• Employed two distinct data-driven computational approaches using computer-generated faces and real photographs.
• Validated the generated face models for perceived criminality and remorse.
• Correlated findings with established models of facial characteristics like dominance, threat, and trustworthiness.

Main Results:

• Successfully generated and validated computational models representing perceived criminality and remorse.
• Convergent findings were obtained from both computer-generated and photographic face approaches.
• Established correlations between criminal/remorseful appearance models and other facial trait models.

Conclusions:

• The developed face models enhance understanding of how facial appearance influences perceptions of criminality and remorse.
• These models provide a valuable tool for researching the impact of perceived criminality and remorse on legal decision-making.
• Findings can inform the development of legal policies to mitigate biases related to facial perceptions.