By Ara Sarafian, The Conversation
A study, published in Current Biology today, found the computer system can “see” distinctive features of facial expressions that are not noticed by people.
Back in 2005, an Australian study found that 4% of medical or psychiatric cases were reported to involve symptom exaggeration or probable symptom exaggeration.
The new research found the accuracy with which humans discerned faked expressions was no greater than that of random chance. With training, their accuracy improved to 55%.
The computer system uses pattern recognition to achieve an 85% accuracy.
The mouth is the greatest predictor of falsified expression: how and when it opens. The researchers found that people who are faking open their mouth too regularly and with less variation.
It’s an interesting development in pain diagnosis but has been given a cautious welcome by Australian researchers.
Social psychologist Lisa Williams, from the University of NSW, studies emotional responses and said if a patient was giving an exaggerated wince or sign of pain then it was important to find out why.
“There are cases in which the reasons might be broader than the immediate context. For example, an employee who might fake abdominal pain to be excused from work,“ she said.
“This could reflect a simple desire to leave work that day, but it also might be driven by the need to be home to handle a family situation, or the desire to get away from an unpleasant colleague.
“In such cases, it would be important not only to address the deeper reason, but also to figure out why the person felt the need to fake pain rather than be upfront about the ultimate cause.”
Measuring pain is often a subjective process, involving verbal and numeric self-rating scales (being asked to rate your pain on a scale of 1 to 10), behavioural observation scales and physiologic responses.
“What is 8/10 on a pain scale for one person is bound to be different to 8/10 for another person,“ Dr Williams said.
“And 8/10 on Monday may not be 8/10 on Friday. If computers were able to assess pain, then we would gain a more objective tool.”
But she said recent research had shown that not all people express pain on their faces.
“Some individuals are simply less facially expressive than others,” she said. “It would be problematic if a computer algorithm determined that someone low in expression was not in pain, simply because there was little or no indication on the face.”
Clinical psychologist Simon Crisp, from Deakin University, said the popular cliche was of people faking pain to get out of work. But the opposite happened too.
“At times, people might want to appear to be well when they’re not – to not disappoint or to meet expectations,” he said.
“For example, when someone needs to do a conference presentation, they might want to look healthy and well when it might not be the case.”
Beating the machine
Dr Crisp said such any computer diagnosis system for pain should be used with caution. He likened it to a lie detector, which is often relied upon to make definitive decisions.
“The technology might indicate someone is not experiencing any pain when in fact they are; for others, it might show that they’re in extreme pain when they’re not, but in fact are highly expressive,“ he said. “This might incorrectly suggest that the person is not genuine.”
Dr Williams said it’s important to use multiple sources of information to determine someone’s true level of pain.
“Using a computer algorithm alone has the potential to produce misleading judgements,” she said.
“It’s not that this endeavour doesn’t carry great promise, it’s that it also is risky given what we know about the lack of one-to-one correspondence between facial expressions and internal states.”
James McAuley, from Neuroscience Research Australia, is an expert in pain and he agrees.
“Facial expression is the communication of pain; it doesn’t describe, and it isn’t required for the experience of pain,” he said.
“These are very interesting technological developments but we are a long way from understanding, or even identifying, the subjective experience of pain using these methods.
“Pain is far too complicated for that.”