Publikation

A Software for Identifying the Facial Features Driving the Perception of Personality Traits

Anke Hirsch, Fabrizio Nunnari, Alexis Heloir

In: 4th Workshop on Virtual Social Interaction. Workshop on Virtual Social Interaction (VSI) 4th December 17-18 London United Kingdom 12/2018.

Abstrakt

The appearance of a person influences the perception of his personality. People use static as well as dynamic cues to assess others. The clothes, facial expression, and posture of a person influence the viewer's perception of its personality; and the same applies to a virtual character. In this context, it would be helpful for a designer to have the tools to shape a virtual character so that its physical aspect conveys the intended personality traits to the observer. In order to run this category of experiments, our research units are building a new open-source platform called DeEvA (https://github.com/fnunnari/deeva/). To test and validate the platform, we recently conducted an experiment to detect which facial features of a virtual character drive the perception of the personality composed of five psychological constructs: dominance, trustworthiness, agreeableness, attractiveness, and narcissism. For each of the five traits separately, we ran a multivariate logit regression with the purpose to find the best predictors of a trait score from the values of the physical attributes.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence