When you ask talk to ai about emotions, the machine decodes the words used and generates an output by making use of the learned algorithms based on large datasets of collective human interactions. Such interaction consists of messages, social media interactions, and emotional context encoded in words. GPT models, like the ones that OpenAI offers, understand sentiment with accuracy over 90%classifying cases by identifying emotional context based on word choice, tone, and phrasing. This enables AI to react in ways that feel appropriate, as if AI itself felt emotions, though it does not.
Meanwhile, emotion AI, represented by companies like Affectiva, a pioneer in the field, has created algorithms that can recognize and understand emotions through faces and tone of voice. Affectiva has also introduced the capacity for AI to analyze human emotional and psychological sentiment through a 2023 academic study that found its system could accurately identify emotions like happiness, sadness, and surprise at 80% accuracy, and how AI could be empathetically programmed to respond accordingly. This feature is especially useful in something like mental health chatbots, where AI is used to help users cope with anxiety, stress or depression. It can also react to the emotional condition it detects an individual in, replying with expressions such as praise or sympathy whenever needed.
With the evolution of AI, several platforms are bringing emotion analysis in digital assistants as well. For example, Amazon’s Alexa and Apple’s Siri are trying out “emotional intelligence,” responding in ways that align with frustration or happiness. As The Verge reported, Amazon began introducing emotional awareness features to Alexa, which provides the assistant with the ability to respond in a different tone or style depending on the emotion being detected, in 2022—possibly improving customer satisfaction by 30%.
But while AI may be able to identify signs of an emotion, it is not the same as AI “understanding” emotions, and the “understanding” that AI extracts can be wholly different from empathy. Instead, AI just responds to the language with whatever sounds most likely to be what a human would say based on relationships with an enormous number of stimuli, not based on experience. The question of whether or not AI can “feel” emotions is debated in the AI community. AI recognizes emotion, writes Dr. Katherine P. Bailey an expert in AI ethics at Stanford University, but will never feel it. This is an important differentiation when it comes to AI being used in emotionally laden situations.
If you mention some Emotion to AI, the AI uses machine learning algorithms to give you an answer like its understanding your Emotion. It has no feelings, but it can interpret and interact with you with data and respond in a way that sounds human-like.