AI has serious defects in recognizing human emotions. How to solve this problem?

As artificial intelligence is used to make more decisions about our lives, engineers have found ways to make it more emotional and intelligent. This means that AI can identify some emotions naturally generated by human beings. The most obvious thing is to scan a person's face and know their feelings at that time.

To achieve this, technology companies such as Microsoft, IBM and Amazon all sell their so-called "emotion recognition" algorithm, which infers people's feelings based on facial analysis. For example, if someone is frowning and his lips are tight, it means that he is angry. If their eyes are wide open, their eyebrows are high and their mouths are wide open, it means they are afraid, and so on.

Customers can use this technology in a variety of ways, from automatic monitoring systems that look for "anger" threats to job interview software that promises to eliminate boring and uninterested job seekers. But how to treat people's feelings is controversial. A recent study shows that these algorithms have no solid scientific foundation.

The research was commissioned by the Psychological Science Association, and five outstanding scientists in this field carefully examined various algorithms and evidence. They spent two years examining these data and reviewed more than 1000 different studies. The basic conclusion of this study is that emotions are expressed in many ways, which makes it difficult to reliably infer someone's feelings from a simple set of facial movements.

Studies have found that people don't frown when they are angry more than 70% of the time. Most importantly, they often frown when they are not angry. This means that companies that use artificial intelligence to evaluate people's emotions will mislead consumers. Studies have shown that some studies that have a strong correlation between facial expressions and emotions often have methodological defects.

The core appeal of artificial intelligence is to let machines simulate the information process of human consciousness and thinking, and can respond in a way similar to human intelligence. It is necessary to examine artificial intelligence from the dimensions of human IQ and EQ. At present, the research on artificial intelligence is mostly limited to pattern recognition that tends to the direction of IQ. Through the comprehensive application of natural language processing, machine learning, pattern recognition, thing perception, logical reasoning and other technologies. The machine has certain logical thinking and judgment ability. However, at present, artificial intelligence emotion recognition still faces the difficulty of understanding and expressing complex emotions without human intervention.

Emotional historian Thomas Dixon is very optimistic about this. He believes that for AI, the mechanism of emotion generation may be similar to the weighted calculation of many factors extracted by the brain. As long as the weight coefficients of various factors in different cultures and individuals are clear, the problem of emotion recognition will be solved.