New research has highlighted how deepfake technology can be used to investigate the effects of facial expressiveness on social outcomes, particularly in job interviews. The study, published in Scientific Reports, explores how facial behaviours like gazing, smiling, and nodding influence perceptions of competence, warmth, and overall favourability during first impressions.

Incorporating AI-generated deepfake videos, the study replicated prior findings, which suggest that people who are more facially expressive tend to be perceived more positively. Using deepfake technology, the researchers standardised facial behaviours across different participants, eliminating biases linked to gender, age, or ethnicity. This allowed for a more controlled analysis of how nonverbal cues affect social judgements.

The experiment involved 159 individuals whose facial behaviours were manipulated through AI. Their videos were presented to a group of over 800 observers, who rated the subjects on competence, warmth, and overall impression. The results revealed a clear pattern: expressive individuals were consistently rated higher across all dimensions than those who appeared less engaged or expressive.

Facial expressiveness, which includes gazing directly at the interviewer, nodding, and smiling, has long been associated with positive interpersonal outcomes. But variations in participants’ natural expressiveness limited earlier studies. Deepfake technology allowed the researchers to bypass these limitations, offering a consistent level of expressiveness across all test subjects. This approach provided valuable insights into how subtle variations in nonverbal behaviour can dramatically alter social perceptions, especially in a high-stakes environment like a job interview.

The study also investigated whether cultural or gender differences played a role in how expressiveness was perceived. While expressiveness had a positive effect across all groups, non-expressive targets from different cultural backgrounds, particularly Indian participants, were perceived more favourably than their Swiss counterparts. Interestingly, gender also played a role, as women, particularly from India, were rated more competent than their male or Swiss female counterparts when not showing expressive behaviours.

This research highlights the potential of deepfake technology not just for entertainment or malicious use, but also for scientific applications. By standardising nonverbal behaviour, it opens new doors for understanding how subtle nonverbal cues impact important social outcomes. For job candidates, the study reinforces the importance of appearing engaged during an interview—not just through verbal communication, but also through facial expressiveness.

In the broader context, the use of AI-generated characters in psychological and behavioural research could help control for variables that are often difficult to manage in traditional studies. This could lead to more reliable findings on the influence of nonverbal communication across different cultural and gender lines.