Cross-cultural judgments of personality and decision-making in job interviews

Abstract

Introduction. Humans express their personality through their every-day behaviour. In this way, they provide other humans with cues for detecting personality with a certain degree of accuracy. Accurately judging a stranger’s personality after short interaction is a useful skill with implications for applied settings such as personnel selection, management, and negotiation. Amongst the cues that a judge uses to evaluate a target, Nonverbal Behaviours (NVBs) have a prominent role because of their immediacy, pervasiveness, and spontaneity. Research in the area, which has thrived in the last three decades, is called to focus on two contemporary and relevant trends: first, the emergency of algorithms that can identify and combine humans’ NVBs to make artificial judgments of personality; second the need to understand nonverbal expression and detection of personality across cultures in a growingly multipolar world. Aims. The research has two primary aims: a) to promote the mutual understanding of the traditional psychological discipline of nonverbal personality judgments and the more recent work in personality computing; b) to test if exists a same-culture advantage in the accuracy of personality judgments and to explore the role that nonverbal behaviours play in cross-cultural personality judgments. Although both the aims are primarily theoretical, we have also considered, in collaboration with Aon Assessment Solutions, the applied value of the findings for the employment interview. Methodology. The research is founded on the Brunswik Lens Model (Brunswik, 1956). The model includes four elements that are vital for the study of nonverbal personality judgments: accuracy (the relation between a target personality and its judgments), cue validity (the relation between personality and nonverbal cues), cue utilisation (the relation between nonverbal cues and judgments), and cue sensitivity (the relation between cue validity and cue utilisation). For the meta-analytic study presented in Chapter II, we have conducted a systematic search using Bramer and De Jonge's (2015; 2018) methodology. We aggregated the data using multi-level meta-analysis to estimate each cue's mean validity and utilisation and the mean cue sensitivity. In Chapter IV, we introduced a methodology for the automatic measurement of nonverbal facial data through the software OpenFace 2.0 (Baltrusaitis et al., 2018) and our own R package OpenFaceR. The methodology was also used for measuring nonverbal cues in the applied study presented in Chapter VI. The studies in Chapters V and VI both aimed at identifying same-culture advantage in cross-cultural personality judgments. In both studies, judges from Western and Eastern cultural backgrounds were asked to judge the personality of Western and Eastern targets. Our primary hypotheses were tested through Multilevel Linear Models in Chapter V and through multigroup Structural Equation Models and Two-way repeated measure ANOVA in chapter VI. Results. The meta-analysis in Chapter II showed that nonverbal behaviours are valid predictors of Extraversion (r ranging from .11 to .23) and of Extraversion Judgments (r ranging from .13 to .42). Cue sensitivity was consistently very high (r = .87). Moderation analysis showed an effect of targets’ gender on cue validity (with women producing more valid cues than men), an effect of setting on cue utilisation (cues utilised more when looking at stranger’s pictures rather than videos or live interactions), and an effect of interaction duration on cue validity and cue utilisation (with shorter interaction producing more valid and utilised cues). The studies in Chapters V and VI studied the effect of targets and judges’ culture on personality judgments. Both studies showed a large effect of targets’ culture on the production of NVBs, and a cultural advantage in accuracy when judging personality traits. We found the highest accuracy when Western judges rated Western targets and the lowest accuracy when Western judges rated Asian targets. The study presented in chapter V also showed that judges have high cue sensitivity when rating targets from their own culture. The study in chapter VI also demonstrated a same-culture bias in job interview ratings for Western judges but not for Asian judges. Conclusions. A framework for closer collaboration between Computer Science and Psychology disciplines in studying nonverbal personality detection has been presented. The framework has been used to include software-extracted nonverbal behaviours in our analysis. We determined the existence of a same-culture advantage for personality judgments, and our data suggest that a higher level of cue sensitivity explains this advantage. Beyond our hypothesis, we also found baseline differences in the expression and detection of personality across cultures. A provisional model to explain the accuracy of judgments within and across different cultures is proposed in chapter VII. The research has applications for the employment interview practice, particularly for cross-cultural and automated job interviews. We propose best practices aligned with our findings.