AI systems that use emotional reading algorithms to evaluate facial expressions are not very good at lie detection


Most algorithms have probably never heard the Eagles’ song, “Lyin’ Eyes.” Otherwise, they’d do a better job of recognizing duplicity.

Computers aren’t very good at discerning misrepresentation, and that’s a problem as the technologies are increasingly deployed in society to render decisions that shape public policy, business and people’s lives.

Turns out that algorithms fail basic tests as truth detectors, according to researchers who study theoretical factors of expression and the complexities of reading emotions at the USC Institute for Creative Technologies.

The research team completed a pair of studies using science that undermines popular psychology and AI expression understanding techniques, both of which assume facial expressions reveal what people are thinking.

“Both people and so-called ’emotion reading’ algorithms rely on a folk wisdom that our emotions are written on our face,” said Jonathan Gratch, director for virtual human research at ICT and a professor of computer science at the USC Viterbi School of Engineering.

“This is far from the truth. People smile when they are angry or upset, they mask their true feelings, and many expressions have nothing to do with inner feelings, but reflect conversational or cultural conventions.”

Gratch and colleagues presented the findings today at the 8th International Conference on Affective Computing and Intelligent Interaction in Cambridge, England.

Of course, people know that people can lie with a straight face. Poker players bluff. Job applicants fake interviews. Unfaithful spouses cheat.

And politicians can cheerfully utter false statements.

Yet, algorithms aren’t so good at catching duplicity, even as machines are increasingly deployed to read human emotions and inform life-changing decisions.

For example, the Department of Homeland Security invests in such algorithms to predict potential threats. Some nations use mass surveillance to monitor communications data.

Algorithms are used in focus groups, marketing campaigns, to screen loan applicants or hire people for jobs.

“We’re trying to undermine the folk psychology view that people have that if we could recognize people’s facial expressions, we could tell what they’re thinking,” said Gratch, who is also a professor of psychology.

“Think about how people used polygraphs back in the day to see if people were lying.

There were misuses of the technology then, just like misuses of facial expression technology today. We’re using naïve assumptions about these techniques because there’s no association between expressions and what people are really feeling based on these tests.”

To prove it, Gratch and fellow researchers Su Lei and Rens Hoegen at ICT, along with Brian Parkinson and Danielle Shore at the University of Oxford, examined spontaneous facial expressions in social situations.

In one study, they developed a game that 700 people played for money and then captured how people’s expressions impacted their decisions and how much they earned. Next, they allowed subjects to review their behavior and provide insights into how they were using expressions to gain advantage and if their expressions matched their feelings.

This shows a woman's face and binary code

The scientists found that smiles were the only expressions consistently provoked, regardless of the reward or fairness of outcomes. Additionally, participants were fairly inaccurate in perceiving facial emotion and particularly poor at recognizing when expressions were regulated. The findings show people smile for lots of reasons, not just happiness, a context important in the evaluation of facial expressions. The image is in the public domain.

Using several novel approaches, the team examined the relationships between spontaneous facial expressions and key events during the game.

They adopted a technique from psychophysiology called “event-related potentials” to address the extreme variability in facial expressions and used computer vision techniques to analyze those expressions.

To represent facial movements, they used a recently proposed method called facial factors, which captures many nuances of facial expressions without the difficulties modern analysis techniques provide.

The scientists found that smiles were the only expressions consistently provoked, regardless of the reward or fairness of outcomes.

Additionally, participants were fairly inaccurate in perceiving facial emotion and particularly poor at recognizing when expressions were regulated.

The findings show people smile for lots of reasons, not just happiness, a context important in the evaluation of facial expressions.

“These discoveries emphasize the limits of technology use to predict feelings and intentions,” Gratch said. “When companies and governments claim these capabilities, the buyer should beware because often these techniques have simplistic assumptions built into them that have not been tested scientifically.”

Prior research shows that people will make conclusions about other’s intentions and likely actions simply based off of the other’s expressions.

While past studies exist using automatic expression analysis to make inferences, such as boredom, depression and rapport, less is known about the extent to which perceptions of expression are accurate. These recent findings highlight the importance of contextual information when reading other’s emotions and support the view that facial expressions communicate more than we might believe.

Funding: The research papers were funded by the U.S Army and a grant (#FA9550-18-1-0060) from the European Office of Aerospace Research and Development.

Tech companies like Microsoft, IBM, and Amazon all sell what they call “emotion recognition” algorithms, which infer how people feel based on facial analysis.

For example, if someone has a furrowed brow and pursed lips, it means they’re angry.

If their eyes are wide, their eyebrows are raised, and their mouth is stretched, it means they’re afraid, and so on.

A review was commissioned by the Association for Psychological Science, and five distinguished scientists from the field were asked to scrutinize the evidence.

Each reviewer represented different theoretical camps in the world of emotion science. “We weren’t sure if we would be able to come to a consensus over the data, but we did,” Barrett says. It took them two years to examine the data, with the review looking at more than 1,000 different studies.

Their findings are detailed — they can be read in full here — but the basic summary is that emotions are expressed in a huge variety of ways, which makes it hard to reliably infer how someone feels from a simple set of facial movements.

“People, on average, the data show, scowl less than 30 percent of the time when they’re angry,” says Barrett.

“So scowls are not the expression of anger; they’re an expression of anger — one among many. That means that more than 70 percent of the time, people do not scowl when they’re angry. And on top of that, they scowl often when they’re notangry.”

This, in turn, means companies that use AI to evaluate people’s emotions in this way are misleading consumers. “Would you really want outcomes being determined on this basis?” says Barrett. “Would you want that in a court of law, or a hiring situation, or a medical diagnosis, or at the airport … where an algorithm is accurate only 30 percent of the time?”

The review doesn’t deny that common or “prototypical” facial expressions might exist, of course, nor that our belief in the communicative power of facial expressions plays a huge role in society. (Don’t forget that when we see people in person, we have so much more information about the context of their emotions than simplistic facial analysis.)

The review recognizes that there’s a huge variety of beliefs in the field of emotion studies. What it rebuts, specifically, is this idea of reliably “fingerprinting” emotion through expression, which is a theory that has its roots in the work of psychologist Paul Ekman from the 1960s (and which Ekman has developed since).

Studies that seem to show a strong correlation between certain facial expressions and emotions are often methodologically flawed, says the review. For example, they use actors pulling exaggerated faces as their starting point for what emotions “look” like. And when test subjects are asked to label these expressions, they’re often asked to choose from a limited selection of emotions, which pushes them toward a certain consensus.

When people are asked to label emotions on faces and aren’t given a set of choices, their answers vary considerably, as this chart shows. 

People intuitively understand that emotions are more complex than this, says Barrett. “When I say to people, ‘Sometimes you shout in anger, sometimes you cry in anger, sometimes you laugh, and sometimes you sit silently and plan the demise of your enemies,’ that convinces them,” she says. “I say, ‘Listen, what’s the last time someone won an Academy Award for scowling when they’re angry?’ No one considers that great acting.”

These subtleties, though, are rarely acknowledged by companies selling emotion analysis tools.

In marketing for Microsoft’s algorithms, for example, the company says advances in AI allow its software to “recognize eight core emotional states … based on universal facial expressions that reflect those feelings,” which is the exact claim that this review disproves.

This is not a new criticism, of course. Barrett and others have been warning for years that our model of emotion recognition is too simple.

In response, companies selling these tools often say their analysis is based on more signals than just facial expression.

The difficulty is knowing how these signals are balanced, if at all.

One of the leading companies in the $20 billion emotion recognition market, Affectiva, says it’s experimenting with collecting additional metrics.

Last year, for example, it launched a tool that measures the emotions of drivers by combining face and speech analyses.

Other researchers are looking into metrics like gait analysis and eye tracking.

In a statement, Affectiva CEO and co-founder Rana el Kaliouby said this review was “much in alignment” with the company’s work.

“Like the authors of this paper, we do not like the naivete of the industry, which is fixated on the 6 basic emotions and a prototypic one-to-one mapping of facial expressions to emotional states,” said el Kaliouby.

“The relationship of expressions to emotion is very nuanced, complex and not prototypical.”

Barrett is confident that we will be able to more accurately measure emotions in the future with more sophisticated metrics.

“I absolutely believe it’s possible,” she says. But that won’t necessarily stop the current limited technology from proliferating.

With machine learning, in particular, we often see metrics being used to make decisions — not because they’re reliable, but simply because they can be measured.

This is a technology that excels at finding connections, and this can lead to all sorts of spurious analyses: from scanning babysitters’ social media posts to detect their “attitude” to analyzing corporate transcripts of earnings calls to try to predict stock prices.

Often, the very mention of AI gives an undeserved veneer of credibility.

If emotion recognition becomes common, there’s a danger that we will simply accept it and change our behavior to accommodate its failings.

In the same way that people now act in the knowledge that what they do online will be interpreted by various algorithms (e.g., choosing to not like certain pictures on Instagram because it affects your ads), we might end up performing exaggerated facial expressions because we know how they’ll be interpreted by machines.

That wouldn’t be too different from signaling to other humans.

Barrett says that perhaps the most important takeaway from the review is that we need to think about emotions in a more complex fashion.

The expressions of emotions are varied, complex, and situational. She compares the needed change in thinking to Charles Darwin’s work on the nature of species and how his research overturned a simplistic view of the animal kingdom.

“Darwin recognized that the biological category of a species does not have an essence, it’s a category of highly variable individuals,” says Barrett. “Exactly the same thing is true of emotional categories.”

Media Contacts:
Gary Polakovic – USC
Image Source:
The image is in the public domain.

Original Research: The study will be presented at the 8th International Conference on Affective Computing & Intelligent Interaction (ACII 2019) in Cambridge, UK.


Please enter your comment!
Please enter your name here

Questo sito usa Akismet per ridurre lo spam. Scopri come i tuoi dati vengono elaborati.