Synthetic empathy given by a machine does make sense, yet there is something surreal about it. Even if the other person doesn’t actually “feel” what we feel, we can still feel a sense of comprehension thanks to our brain’s capacity to mirror others’ emotions and our fundamental need for social connection.
When we experience empathy from others, including chatbots, so-called mirror neurons in our brains become active, fostering a sense of connection. Empathy, of course, has many facets, and it’s possible to argue that in order for us to properly experience empathy, another human body in the room sharing our sentiments temporarily is necessary. And this has been vividly shared by Arvind Narayanan, a professor of computer science at Princeton, who set up a speech interface for ChatGPT for his daughter, who is almost four years old. The revelation comes as part of an ongoing experiment.
Clinical psychologist Thomas Ward, at Kings College London who has studied the use of software in therapy, provides a warning about assuming that AI can effectively replace a hole for those who need mental health help, especially if their problems are serious. For example, a chatbot is unlikely to accept that a person’s feelings are difficult to comprehend. In other words, ChatGPT rarely answers with “I don’t know,” as it was built to err on the side of confidence rather than caution.
Additionally, Ward commented on how people should be cautious about frequently using chatbots as emotional outlets. Inconspicuous features of human interaction, such as hand contact or knowing when to speak and when to listen.
Narayanan explained that his daughter was naturally inquisitive and frequently asked questions about creatures, plants, and the human body. He believed ChatGPT might provide her with insightful responses. To his astonishment, once he informed the chatbot that he was chatting with a young child, it performed flawlessly in terms of empathetic behaviour.