Computers trained to recognize emotions

While humans have always been better at detecting and responding to emotions than computers, new research done at MIT is showing that, in some cases, computers are taking the lead over their human counterparts.

The study focuses on the act of smiling, honing in on the different reasons people smile, whether out of happiness and delight, or pure frustration. And using the results from the study for a large sample of people, researchers on this project have fed information to computers, which are actually better at telling the different types of smiles apart.

Experiments involved asking participants to act out expressions associated with specific emotions, which were recorded by a webcam. They then had to fill out a purposely-frustrating form or watch a video made to evoke feelings of delight, and their reactions were recorded as well.

One of the most interesting findings was that the vast majority of those asked to feign frustration did not smile in their forced attempts, but upon experiencing frustration in an unprompted situation, they did. Additionally, there is a difference in the way people smile; those who are delighted tend to have a gradual build-up to the smile, whereas frustrated smiles are quick and fleeting.

The main aim of this study is to help unravel the mysteries of emotions. In particular, those affected by autism may have a difficult time interpreting emotions; while a smile is viewed as a positive thing, this study demonstrates that isn’t always the case. Additionally, those who are public speakers or figures in the spotlight might benefit from better understanding the timing in reactions and how the slightest difference in facial emotions can be interpreted differently.

Leave a Reply

Post Navigation