UC Berkeley study uses GIFs to identify 27 new states of emotion

Alan Cowen/Courtesy

Related Posts

A study led by campus psychology researchers used GIFs to identify 27 new states of human emotion connected on a gradient, opposing the long-established belief that there are a few basic emotional states.

Graduate student Alan Cowen was the driving force of the research, along with campus psychology professor Dacher Keltner. Their study was published Sept. 5 in Proceedings of the National Academy of Sciences of the United States of America.

“There’s so much evocative content on the internet being generated every day, and I wanted to utilize it to study a wider range of human emotion,” Cowen said.

The research involved scanning the internet for various GIFs that would elicit different emotions, using data-scraping techniques. According to the study, more than 800 men and women of varying demographics reacted to 2,185 GIFs. The videos represent psychologically significant situations, such as war, funerals, humor and animals, according to Cowen.

Keltner said the data from their research suggests that there is a general consensus to how people label their emotions. Cowen explained that if one group reported a certain response, such as “adoration,” and another group reported a similar response, such as “joy,” to the same video, the two emotions could be grouped with “happy.”

“We did a mathematical analysis to find out how many dimensions people needed to describe their emotion,” Cowen said. “Turns out there are 27, and they’re on a continuous gradient of emotional states that link together different emotions.”

Scientists previously understood that there were only five to six emotions, but Cowen and Keltner were able to use modern crowdsourcing tools to collect a large amount of data that could quantify and visualize the broad spectrum of human emotion.

There has been some controversy as to whether or not the experience of an emotion is based on the labeling of it, according to Keltner. There is a common assumption that distinct emotions cannot overlap with others, yet Cowen and Keltner’s research suggests that an experience can be a culmination of multiple emotions.

“There’s a critique in literature right now that says if we label our emotions, there’s this feeling that we construct it,” Keltner said. “Maybe what we’re labeling isn’t real.”

The study is significant because it may change the way people orient their environment to their emotions, Cowen said. For example, Netflix offers different genres, such as horror and thriller, based on an intuitive understanding of emotional states, according to Cowen. But Netflix may want to quantify the level of horror that viewers prefer, which Cowen said can be modeled using their research. Companies can use their research to cater to people’s more nuanced emotional states, based on the emotion spectrum they developed.

“This could influence the way we interact with the machines,” Cowen said. “I hope science will use this taxonomy of emotions to explore how emotional states vary across cultures.”

Gioia Von Staden and Carina Zhao at [email protected].