Social Engineering Blogs

An Aggregator for Blogs About Social Engineering and Related Fields

Expressing Corruption?

Many of us worry about corruption amongst political elites, but could it be possible to actually recognize it in their faces?

In a pretty creative study, a team of psychologists from the California Institute of Technology sought to explore whether people could detect evidence of corruption by government officials by providing them with pictures of their faces. This study helps shed light both on efforts to effectively read other people but also on efforts by citizens to better evaluate our elected representatives.

While this may seem initially like a pretty far-fetched idea, there is a long history of research showing that people tend to make competent decisions about people’s trustworthiness from images of their faces. This has even been applied to potential leaders, where prosocial outcomes lead to positive evaluations.

However, this study makes an important break in shifting from just evaluations of a person’s charisma and perceived competence to actually determining if they have been practicing deception. Still, there is some prima facie credibility in that guilty expressions are generally identifiable.

To answer their question, the study authors undertook a series of experimental designs, showing images of politicians and asking participants to identify salient traits, such as corruptibility, dishonesty, and selfishness but also including more prosocial tendencies like competence and ambition.

In the first of these studies, participants were exposed to a series of 72 photos of actual elected officials in the United States. Of those, half had been convicted of some form of corruption, such as violations of campaign finance laws.

Before exposing participants to these photos, they were prompted with an instruction that they would have to designate the official’s level (1-5) of a given trait as quickly as possible, and they only had about four seconds to do so. This sought to ensure that people were judging based on spontaneous and initial reactions of emotional recognition.

Subsequent studies functioned similarly in an effort to strengthen the generalizability of any findings. This included varying the level of government that a given official held or using variants of the initial traits.

Overall, this series of experiments found broad support for the ability of participants to identify political corruption in faces of elected officials. This held up across variants, suggesting that it would apply more broadly outside of a limited experimental setting.

While many questions about generalizability and the precise causal mechanisms remain, this ambitious study does give us further evidence that our ability to read faces and detect deception has great potential even in photographs.

It would of course be interesting to see if an observer professionally trained in deception detection and people reading would do better.

In the meantime, check out some of Dr. Matsumoto’s work on politics and deception!

Prompted to Mimic Faces

Spontaneous Facial Mimicry

At the beginning of each block, an introduction-picture was presented, followed by 2–4 trials. (A) In the Emotion-Inference condition, the instruction “How does XXX (e.g., Hashimoto) feel?” was presented in Japanese. (B) In the Passive (control) condition, a fixation cross was presented. Reprinted from the ATR Facial Expression Image Database DB99 under a CC BY license, with permission from ATR-Promotions Inc., original copyright (2006).

Facial mimicry is known to be central to understanding the emotional states of others, but this exciting new study looks at the conditions under which we engage in such activity.

Emotional recognition is incredibly central to social interactions, and facial mimicry allows us to do so instantaneously. However, there is dispute over when we spontaneously or automatically engage in such behavior. It is this dispute which an exciting new paper in Plos One attempts to answer.

Specifically, the study authors are attempting to determine what factors make an individual engage in mimicry and what factors prevent them from doing so. This builds off of previous research finding that socio-ecological factors, like group membership and the identities of the people in question, inhibit or encourage the spontaneous mimicry necessary for emotional recognition.

While the role of such social factors is an ambitious one, the study authors sought a more manageable effort within that context. Previous research has indicated that facial mimicry occurs when there is an intention on behalf of the individual to deduce emotional states, and they sought to extend those findings in a more robust fashion, with an eye to the broader question.

Participants were recruited from the Hokkaido University student body and asked to review a selection of photographs and videos of people’s faces exhibiting various emotions. During this process, their facial muscles were analyzed to detect the presence of facial mimicry, in a similar fashion as previous studies.

However, not all participants were asked to identify the emotions displayed in the various images. Instead, half of them were simply asked to record age, gender, and other demographic features. This required emotional recognition from only half of the group.

The hypothesis was that those primed to look for emotions would also spontaneously engage in facial mimicry, while those looking at demographic factors would not.

The electromyography (EMG) scans were able to detect facial muscle movement as a reasonable indicator of facial mimicry, and it turns out that those asked to detect emotions showed markedly higher rates of muscle movement than those who were not. This provides pretty compelling evidence in support of their hypothesis.

This is definitely one of those times where many readers may be unclear exactly why any of this matters. Since facial mimicry, and thus emotional recognition, are not automatic but tied to intention, the social context begins to be seen as more important.

Differing group membership, for instance, can actively inhibit the intention to recognize emotions, leading to ignorance or devaluing of emotions. This is particularly fascinating (and troubling) given our previous blogs on the role that group membership has on violence and anger.

What it also means is that you, as an aspiring people reader, can choose to engage in facial mimicry and choose to better read people. This is something we can learn how to do, which is what Humintell tries to do, but it is also something we have to want to do.

 

How Many Faces Can You Recognize?

many-faces-facial-recognition

Who is that person in your office? On your bus? On the television?

Many of us are constantly barraged with different faces, and it can be hard to keep track or even to remember some familiar faces at all! Yet, why is this?

In a recent paper in the Proceedings of the Royal Society, a team of researchers sought to analyze whether we create a sort of list or catalogue of faces that we know. In other words, how many faces are we capable of remembering at any given time? While they find an average of about 5000 remembered faces, individual variation seems to play a huge factor in one’s ability to recognize faces.

Importantly, this is not a paper describing what our memory is capable of knowing. Rather, they are trying to determine how many faces people tend to keep in their working memory. Interestingly, most anthropological research finds that humans tend towards small groups of maybe 100 people, but this must be contrasted with the demands of modern society to recognize a multitude of faces every day.

There are, of course, many types of facial recognition, broadly speaking, that complicate this effort. For instance, we may recognize people’s faces whom we have never met or even never seen in person, or we could fail to recognize someone when seen in a novel context. For precision, this paper sought to look at whether facial recognition held up when seeing a familiar face in such novel contexts.

This was evaluated using an experimental design. Participants were exposed to a series of 3441 public figures and asked which they recognized. These were randomly interspersed with slight variations of those same faces an additional 3441 times, so each face was seen twice. This allowed researchers to see if the face was recalled from a previous exposure.

This is only one type of facial recognition, however, as the researchers had to grapple with the multitude of people that we do see every day and know personally. This was looked at by giving participants clear criteria for what constituted a facial memory and asking them to write detailed self-reports of those whom they know personally, including people that they might just happen to see every day on the bus.

When combining recall rates of famous figures with accounts of people known personally, they relied on statistical methods to derive an average estimate of about 5000 people, though this faced incredible individual variance from about one to ten thousand, depending on the participant.

These same individual differences were present during each attempt at a robustness check. This means that researchers subset participants in different fashions and also that they changed recall measures to less stringent cases. For instance, this involved looking not at whether they recognize both famous faces in a pair but if they recognize either picture.

You might very well ask, what exactly does this teach us about facial recognition? One this it does tell us is that people have incredibly different abilities to recognize faces in such contexts. Some people may be very bad at doing so, for instance.

Still, given how intertwined facial recognition is with emotional recognition, this is not about some innate ability to recognize either one. Instead, it can be trained like any other skill, which is exactly what Humintell is here for!