Trust is a cornerstone of social interactions, guiding our judgments and behaviors in society. Research has shown that we form impressions of trustworthiness from faces in mere moments, but these assessments are made in isolation. New research published in Cognition expands this understanding by exploring how we perceive trustworthiness not just in individual faces but in crowds, revealing our capacity to quickly gauge the average trustworthiness of multiple faces together.
Previous studies have largely focused on the trustworthiness of individual faces, neglecting how we process faces in groups. This gap is significant given that in real life, we often encounter faces not in isolation but as part of a crowd.
This study fills that gap by investigating whether we can perceive an average level of trustworthiness from groups of faces, a process known as ensemble perception. This concept suggests our brain can extract a ‘summary statistic’ of facial features from a crowd, a useful skill in social situations where assessing every individual face is impractical.
“We were interested in this topic because previous research showed how first impressions of trustworthiness influence our social behaviour in one-to-one interactions,” said study author Fiammetta Marini, a PhD student at the School of Psychology at the University of Aberdeen.
“However, in everyday life we do not have only one-to-one interactions, but we often encounter groups of people. Thus, it was interesting for us to understand also how we form impressions of groups, because these impressions can potentially influence our social behaviour towards that group.”
“For example, imagine being alone and walking down a dimly lit road on a foggy night,” Fiammetta explained. “Suddenly, you come across a group of people. Are they dangerous or not? Should you change your way or continue walking in that direction? In this situation, it is crucial to quickly for an impression of the group who you are facing.”
For their study, Fiammetta and her colleagues conducted a series of four experiments.
Experiment 1 set the foundation, engaging thirty-five participants in an adjustment task. They were presented with sets of computer-generated faces, varied in trustworthiness levels, and were tasked with adjusting a test face to match the average trustworthiness of the previously seen group. The faces were displayed in different set sizes to determine whether participants could accurately gauge the average trustworthiness from varying numbers of faces. This experiment tested the basic premise: Can people extract a general impression of trustworthiness from a group?
Participants were adept at adjusting a test face to reflect the mean trustworthiness of a previously seen set, with accuracy surpassing chance levels across all set sizes. This demonstrated a clear capacity to integrate trustworthiness information from multiple faces, establishing a baseline for ensemble perception in this domain. Interestingly, the task became slightly more challenging as the number of faces increased, suggesting a potential upper limit to the number of faces that can be simultaneously processed for trustworthiness.
In Experiment 2, the researchers sought to replicate and extend the findings of Experiment 1 using a different approach—a rating scale task. Thirty-eight participants viewed the same sets of faces but, instead of adjusting a test face, they rated the average trustworthiness on a scale. This subtle shift in methodology aimed to confirm whether the ensemble perception observed in Experiment 1 was robust across different types of tasks, highlighting the participants’ ability to form trustworthiness judgments in more than one way.
Participants consistently rated the average trustworthiness of face sets accurately, affirming that the ensemble perception of trustworthiness is not confined to a specific task setup. This experiment also indicated that participants could abstract trustworthiness information into a coherent summary statistic, regardless of the method of judgment, underscoring the robustness of this cognitive ability.
Experiment 3 introduced an ecological twist by mixing different identities within each set, challenging thirty-six participants to rate the group’s average trustworthiness. This variation aimed to mirror real-life conditions more closely, where groups are rarely homogeneous. By increasing the ecological validity, the researchers tested whether the ability to perceive average trustworthiness was flexible and adaptable to more complex visual scenes.
The researchers found that the introduction of varied identities within a single trial did not deter participants’ ability to accurately perceive average trustworthiness. The results from this experiment highlighted the adaptability and ecological validity of ensemble perception in trustworthiness judgments.
Finally, Experiment 4 examined the temporal dynamics of ensemble perception by presenting faces to forty-seven participants for varying durations—from as brief as 50 ms to as long as 2000 ms. This experiment aimed to discover the minimum amount of time required for accurate ensemble perception of trustworthiness to occur, providing insight into the speed and efficiency of social judgments.
Participants could form accurate ensemble perceptions of trustworthiness even at very brief exposure times (as short as 250 ms). This rapid processing capability underscores the efficiency of social cognition, suggesting that our judgments of trustworthiness in groups are not only quick but also instinctual.
“Interestingly, we found that people form these average impressions very quickly,” Fiammetta told PsyPost. “Indeed, seeing a group of faces for up to 250 ms is enough for people to form group judgments of trustworthiness!”
However, the ability to process trustworthiness information from groups showed a diminishing return at ultra-short exposures (below 250 ms), indicating a threshold below which our ability to integrate trustworthiness information becomes less reliable.
Together, the findings provide evidence that our judgments about groups are not just a collection of individual assessments but a coherent, averaged perception.
“Our research shows that when we encounter a crowd are able to quickly extract the average trustworthiness from a group of faces, without looking at every single face in the group,” Fiammetta said. “This happens because we combine together all the faces’ features into one average face instead of meticulously examining each single face in the group.”
“This visual mechanism, called ‘ensemble perception,’ is particularly useful given that it permits us to quickly extract social information, such as the overall threat level or approachability and trustworthiness of a group, in situations in which we are required to take fast social decisions.”
While the study offers new insights, it has some limitations, such as the use of computer-generated faces and the focus on male faces to control for gender effects. “This study used computer-generated faces that varied in their level of trustworthiness, as they were controlled stimuli that minimized individual differences,” Fiammetta noted. “However, future work might wish to investigate trustworthiness group impressions also by using naturalistic face images.”
Future work could also examine the practical implications of ensemble perception.
“So far, we don’t know the impact of group impressions of trustworthiness on our social behavior, and it would be very relevant in the future to address this important topic. For example, in situations such as simultaneous police line-ups where multiple suspects are seen all together, if the perceived trustworthiness of a suspect face is affected by the other surrounding faces as we form an average of the group, this could eventually hinder the eyewitness identification of the suspect.”
“We are prone to form impressions of trustworthiness because there is a strong relationship between emotional expressions and social personality attributions from faces, and this relationship extends also to emotionally neutral faces,” Fiammetta added. “Therefore, faces with a neutral expression that subtly resemble a positive emotional expression, in particular with high eyebrows with a surprised-looking shape and an upward shaped mouth, are perceived as trustworthy.”
“However, it’s important to remember that the link between facial features and an individual’s trustworthiness is far from perfect, and we discourage people from using trustworthiness impressions as reliable cues of others’ personality traits. Indeed, even if we form these impressions in a few milliseconds and they impact our social interactions everyday, they are far from accurate and do not reflect whether a person is actually trustworthy or not.”
The study, “Three’s a crowd: Fast ensemble perception of first impressions of trustworthiness,” was authored by Fiammetta Marini, Clare A.M. Sutherland, Bārbala Ostrovska, and Mauro Manassi.
