Construção e evidências de validade de um banco de faces para reconhecimento de expressões faciais de emoções básicas na população brasileira

Abstract

The Recognition of Facial Expressions of Emotions (RFEE) is an essential skill for good social functioning, as it allows recognizing the emotions and possible intentions of the other. An important variable to be considered in the RFEE skill is the cultural context, which tends to modulate this recognition. Thus, the objective of the present study is to build and seek evidence of validity for a series of stimuli of facial expressions of emotions of Brazilian individuals, considering the population distribution of the country according to age group, gender and race. For this, a three-phase procedure was constructed to elicit basic emotions based on studies that also built face databases: 1) Presentation of specific stimuli from the International Affective Picture System (IAPS); 2) Presentation of model photographs from the Pictures of Facial Affect (POFA) database and 3) Specific hypothetical situations taken from a previous study and tested for the present study. Each image was analysed by expert judges in studies with emotions and only faces with 100% agreement between them for the type of emotion expressed by the models were included in the final set. Subsequently, 459 Brazilians of both sexes and different ages and races recognized these faces in the validity evidence analysis stage. The final database consists of 56 high-quality color images of young, adult and elderly individuals (with and without acting experience) and divided according to the country's population distribution for the variables age, gender and race. The final stimuli show good agreement between the expressions perceived by the evaluators and the models' expressions (M=81.6%; SD=7.69), in addition to satisfactory internal consistency with Cronbach's Alpha of 0.70 and McDonald's Omega of 0.68. However, 11 faces showed <70% recognition accuracy, suggesting that the present set needs reapplication in more representative samples. The percentage of correct answers for each item in the database, as well as the characteristics of the models, such as having or not experience in acting, age, gender and race, are also presented so that future studies can choose the most appropriate images to meet their own needs. We hope that the built image set can be used for studies that evaluate the recognition of emotions in the Brazilian context.