Classifier

Background:

Affective computing advancements have led numerous companies worldwide to focus on developing facial emotion recognition systems to enable artificial intelligence to predict human behavior. Major tech giants such as Amazon, Microsoft, and Google provide basic emotion analysis capabilities, while smaller enterprises like Affectiva and HireVue specialize in tailored applications for specific industries such as automotive, advertising, mental healthcare, and recruitment. Even MIT has dedicated affective computing groups striving to develop and assess innovative ways of integrating Emotion AI and other affective technologies to enhance people's lives.

Despite the existence of numerous articles highlighting the dangers and biases associated with facial recognition systems, many companies continue to employ them. This is particularly noticeable during the pandemic, as software like 4 Little Trees generates detailed reports on students' emotional states for teachers and even assesses motivation and focus. However, it is important to recognize that "It is not possible to confidently infer happiness from a smile, anger from a scowl, or sadness from a frown."

In a research paper by Barrett, Adolphs, Marsella, Martinez, and Pollak (2019), which reviews over 1,000 studies conducted over a span of two years, scientists from the Association for Psychological Science identified a wide array of expressions used to convey emotions. This variability makes it challenging to reliably infer someone's emotional state based on a limited set of facial movements. The paper highlights three significant limitations in scientific research that contribute to a general misunderstanding of how emotions are expressed and perceived through facial movements, thereby limiting the applicability of this scientific evidence: limited reliability, lack of specificity, and limited generalizability.