Face perception research offers broad and compelling insights, with many useful (yet potentially worrying) applications. From inferences about personality (Walker & Vetter, Reference Walker and Vetter2016) to the growing use of facial recognition systems in crowd monitoring (Veltmeijer et al., Reference Veltmeijer, Gerritsen and Hindriks2021) and the hiring process (Suen et al., Reference Suen, Hung and Lin2019), this work is here to stay. Therefore, it is extremely important that face perception research guards against harmful biases through theoretical rigor, responsible development, and prudent implementation. To serve as an example, I comment on (1) the continued (and outdated) use of gender as a binary factor in face perception research (see Shen & Shoda, Reference Shen and Shoda2021, as a recent example), (2) the need to further explore the value of age diversity (see Tuncdogan et al., Reference Tuncdogan, Acar and Stam2017, for a discussion of existing and future work), and (3) the importance of expanding theoretical horizons (rather than the excessive “gap-spotting” outlined in Sandberg & Alvesson, Reference Sandberg and Alvesson2011). In short, facial recognition systems need to be based on state-of-the-art social science research, not antiquated (and discrimination-inducing) output.
Here are three opportunities for safeguarding against misguided research and practice:
1. Moving past binary gender
When I decided to use faces, it was not because I was necessarily interested in face perception research. Instead, my colleagues and I realized it was a clever way to investigate the implicit biases that followers hold regarding nonbinary aspects of a leader’s gender (e.g., facial masculinity versus femininity; see Little et al., Reference Little, Burriss, Jones and Roberts2007; Spisak, Homan et al., Reference Spisak, Homan, Grabo and Van Vugt2012). My hypothesis then, and now, is that follower decision-making regarding gender is more sensitive than men versus women. We clearly vary beyond assigned biological sex, and there are important nonbinary signals (or at least cues) influencing leader selection that should not go overlooked—especially given the growing appreciation that a binary understanding of gender in psychological research is an outdated concept (Hyde et al., Reference Hyde, Bigler, Joel, Tate and van Anders2019). Instead, face perception research should explore new spaces, such as how voting preferences are changing now that society is starting to appreciate the realities of gender. A new world is emerging that is far more sensitive to differences in assigned sex, gender, and gender identities. Social science research needs to help facial recognition systems keep up with this liberating time (see Scheuerman et al., Reference Scheuerman, Paul and Brubaker2019).
2. Considering the value of age diversity
Though modern research finds fluctuations in leader preferences based on age and life span, much of the foundation is atheoretical or not entirely representative of reality. Shen and Shoda (Reference Shen and Shoda2021), for instance, found that the likelihood of voting for a candidate decreases as the candidate’s perceived age increases—noting that this negative relationship starts for male candidates around a perceived age of 45. However, this finding stands in stark contrast with almost all congressional and parliamentary profiles around the world (e.g., the average age of the current U.S. House of Representatives is 58.4 and the U.S. Senate is 64.3; see Manning, Reference Manning2021). In reality, younger candidates have the hardest time gaining status and leadership roles (Inter-Parliamentary Union, 2021). It is, therefore, important to explore this youth barrier for insights and opportunities—for example, previous work has found that younger leaders are preferred for establishing and maintaining peace and cooperation (Spisak, Reference Spisak2012) as well as leading change (Spisak et al., Reference Spisak, Grabo, Arvey and Van Vugt2014).
The next generation of face perception research needs to build on such findings to better understand and activate the value of age diversity in leadership and society. Social science, for example, can inform facial recognition systems using age to identify leaders who are more likely to encourage the reduction of fossil fuels and the push toward renewable alternatives (see the “green leadership” example in Spisak et al., Reference Spisak, Grabo, Arvey and Van Vugt2014).
3. Contributing to better practice and policy
The use of facial recognition technology is developing rapidly, even though its reliability and validity is a perennial topic of debate. From the U.S. Transportation Security Administration’s application of “observation techniques” based on the work of Ekman and colleagues (see Heaven, Reference Heaven2020) to recent publications suggesting that emerging facial recognition technology is based on outdated psychological science (e.g., Barrett et al., Reference Barrett, Adolphs, Marsella, Martinez and Pollak2019), the foundation is shaky. Yet, despite this uncertainty (and explicit flaws in some cases), the public and private sectors are pushing forward with automated systems for many tasks, such as assessing personality, monitoring (crowd) emotions, and evaluating job applicants (e.g., Suen et al., Reference Suen, Hung and Lin2019; Veltmeijer et al., Reference Veltmeijer, Gerritsen and Hindriks2021). Vendors often use terms like “scientifically validated” to reassure investors, clients, and the public that any costs associated with bias, discrimination, and inequality are minimal (Harlan & Schnuck, Reference Harlan and Schnuck2021). Also, facial recognition systems are advancing (whether society likes it or not) because they are relevant in many domains and profitable in practice. Therefore, these AI systems must be monitored and shaped so they can become human centered and uplifting, not biased and draconian.
Conclusion
The ubiquity of facial recognition calls for rigorous, multidisciplinary research to ensure the implementation of robust face perception insights. Researchers and practitioners need to thoroughly explore relevant literature, move beyond the binary, and investigate the value of diversity. Otherwise, “scientifically validated” technology will continue to incorporate false assumptions. As social scientists, we must safeguard against this unreliable and invalid future.
