Arrival of Emotion-Detecting AIs – The Ups and Downs of Technology Interpreting Human Emotions

5 mins read

The Birth of Emotion-Detecting AIs and Affectiva’s Auto AI Platform

Recently, multiple software companies have created technology capable of reading and interpreting human emotions. With the birth of emotion-detecting AIs, these computers could now assist professions that require deciphering people’s actions and behaviors by understanding their inner sentiments. It may be relieving to hear the news at first, but how exactly accurate are these AI for measuring one’s emotions? And how should we feel knowing that emotion-detecting software exists to read our inner feelings no matter how we hide them?

Software company Affectiva is one of the few corporations that created systems that are adept at reading and detecting human sentiments. According to Affectiva’s co-founder Rana El Kaliouby, she got inspired by her husband to build an emotionally intelligent computer that can help process and relay one’s feelings to other people even when apart. When she moved from Egypt to the UK for her studies 20 years ago, she left her spouse behind and only communicated through text messages and video calls. Kaliouby realized how blind and neutral technology is toward human emotions, and it became evident when she had a virtual chat conversation with her husband. Even though she texted him that she’s fine, Kaliouby does well hiding her real emotion of stress through a neutral text sent online. 

Affectiva built the Auto AI platform, an emotion-detecting AI that is capable of recognizing emotions like anger, joy, and frustration. The said software is mostly available for use in vehicles, where its installed microphones and cameras could pick on a passenger’s drowsiness – useful for avoiding car accidents due to a driver’s drunken or unstable emotional state. Moreover, a smart assistant changes its speaking tone depending on a passenger’s current temperament, such as talking in a gentle voice when dealing with a frustrated individual. The Auto AI system could prove to become an adaptive vehicle experience on-the-go, but it could also become inaccurate with its interpretation of human emotions. For instance, elderly passengers’ facial expressions might get mistaken by the AI as having driver fatigue when, in fact, the system decoded their sentiments less accurately due to old age’s effects on the human body. 

Using AI to Analyze Emotions – Is it Accurate or Biased?

A “mind-reading machine,” as Kaliouby calls it, deems useful and beneficial during the ongoing pandemic crisis. With family members, friends, and co-workers away from one another, it becomes increasingly challenging to initiate heartfelt talks with them through the other side of the computer screens. As much as it can benefit people to understand and sympathize with others’ emotions, do emotion-detecting AI’s bring more good than harm or vice versa?

Emotional AI is prone to subject itself to bias due to how subjective human sentiments are by nature. One research study discovered that temperamental analysis technology brings more negative emotions to particular human races than others. For instance, emotion-detecting AIs consistent in analyzing employees’ negative feelings in the workplace could affect those individuals’ career progression. Not being able to conceal your inner stress at work could also bring detrimental impacts to your colleagues’ to think and feel the same way.

To prevent bias from completely taking over and destroying the accuracy of emotion-detecting AIs, software companies need to become more observant when training their emotional AIs in decoding human expressions. When teaching AIs how to measure different emotional states in real-time, developers also need to use live data for additional references. For instance, a study revealed that only six out of 19 varying types of smiles could occur to express their emotions when people experience good moments. Human beings also smile to hide their embarrassment, pain, or anxiety. 

The previously mentioned example stresses the requirement of taking other emotional factors into play instead of only relying on historical data. That way, emotion-detecting AIs can do a better job of detecting and interpreting emotions to help people understand themselves and others.