AI: how Darwin laid the foundations of emotion recognition – fode.ca

In 1872, 13 years after his major work on “The Origin of Species”, Charles Darwin published the book “The Expression of Emotions in Man and Animals”.

The central thesis of this book was no less revolutionary than his theory of evolution: the naturalist postulated that humans inherited their emotions from the first animal ancestors, which means that emotions are, so to speak, evolutionary vestiges. Hate, anger, but also love, joy or shame would be evoked by old parts of our nervous system in response to specific triggers and would have a specific and unified expression, a sort of fingerprint.

These photos from Darwin’s book on emotions aim to illustrate contempt, discontent and rejection.

(Image: The expression of emotions in man and animals / Oscar Gustave Rejlander)

The book became a bestseller, but was controversial. For after Darwin had already depicted man as an ordinary species and not as the crowning achievement of creation in his first book, he now attributes emotions to animals – which should not differ significantly from those of humans. His book was therefore ignored by scientific circles and fell into oblivion. It was not rediscovered until the 1960s.

Psychologist Paul Ekman in particular championed Darwin’s theory. He wanted to develop technical methods to make emotions measurable. Ekman postulated basic emotions such as joy, anger, disgust, fear, contempt, sadness, and surprise. In Darwin’s mind, he also believed that these emotions act directly on the movement of the so-called units of action – the muscle groups of the face. Secondary emotions such as shame, love, jealousy would form from a superposition of basic emotions.

The model was popular with scientists who wanted to recognize emotions from video or photos – especially with the rise of artificial intelligence. A few years ago, however, resistance formed: in July 2019, for example, psychologists published a meta-study on the expression of emotions in the journal of the Association for Psychological Science. They also included people from different cultures or infants and toddlers. Their conclusion: People showed “much higher variability” in emotional expression than previously postulated. There is “insufficient scientific evidence” for the claim that one can “reliably infer basic emotional states” of a person from facial expressions alone. Their conclusion: Emotion recognition technology, which was also used by US border guards, for example, is “pseudoscience” that has no basis.

However, the scientific debate surrounding the nature of human emotions and whether they can actually be measured is not unequivocally resolved. However, under pressure from critical scientists such as the AI ​​Now Institute, privacy activists and an increasingly wary public, the developers of this technology are coming under increasing pressure. Microsoft, for example, has completely stopped selling the corresponding software. The use of facial recognition and the use of emotion recognition have been banned in various US states and Canada.

(jle)

To the home page

Devoted beer advocate. Alcohol specialist. General tv evangelist. Zombie practitioner. Travel enthusiast. Avid gamer. Bacon expert. Pop culture fan.

AI: how Darwin laid the foundations of emotion recognition