Emotional intelligence is a hot topic. While soft skills are highly sought after by employers, our technology has not yet caught up. As AI advances, the focus is sharply fixed on how it can establish a real connection to it’s users. “Ok Google” and “Hey Siri” echo through households the world over to help with day to day tasks. People are even turning to AI therapists to offload their innermost anxieties. We are coming into contact with these programmes more than ever before, so the question remains – how do we make conversationalists out of these bots?
Elementary, my dear Alexa – empathy. A cornerstone of human connection, empathy is a broad tapestry comprised of many emotional variables. The Cambridge dictionary defines empathy as:
” the ability to share someone else’s feelings or experiences by imagining what it would be like to be in that person’s situation”
Psychologist Carl Rogers, went a little deeper describing empathy as:
“entering the private perceptual world of the other…being sensitive, moment to moment to the changing felt meaning which flow in this other person… assuming a non judgemental stance, and being careful not to uncover meaning that the other would find threatening”
Empathy requires a deep understanding of a person’s background, culture, past experience and human emotion. It helps us to react and respond with tact, consideration and warmth. An inherently human phenomenon, digital assistants will never truly be able to feel empathy. But with increasing demands for compassionate tech to combat loneliness, enhance customer experience and help us navigate our lives, how does one foster empathy through AI?
Emotional recognition is key to eliciting empathic responses via machine learning. In order to gauge the correct response, AI must clearly identify markers of human emotion, which is no mean feat. Creating the neural networks necessary to respond appropriately requires multiple inputs and broad data sets. Programmes must be able to recognise verbal patterns, learn the appropriate cues and line up the right responses. Given the vast range of human emotion and experience, this process can never be 100% accurate. Day to day, people misread each other’s signals all the time, due to differing communication styles, moods and contexts. Trying to navigate the many aspects of verbal and non verbal communication and lining up the appropriate reaction is challenging for those working in AI research. But developers have come a long way in mimicking empathy within AI software.
Mental health apps are an excellent example of empathic bot communication. Woebot, a digital therapist has been helping patients process their feelings using an AI programme and guides users through a series of Cognitive Behavioural Therapy techniques, designed to improve mood. The app helps users to challenge unhealthy thought patterns and has been found to alleviate symptoms of anxiety and depression after just two weeks. These results are incredible and point to a very real need to expand research in this area to best support its user base. Many mental health apps are free, making these tools essential for those who cannot afford alternative mental health treatment. Furthermore, they are an excellent way to augment and support ongoing mental health care.
Empathy in a mental health sphere can be quite easily intimated, and programmes such as Woebot will have a comprehensive bank of “call and response” replies in their arsenal to echo the emotions of users. Users have come to Woebot specifically to process their feelings and as such, programmers can follow on with suitable pre-programmed replies.
In a recent presentation Casey Sackett, developer at Woebot outlined several methods the company used to get to know their users and share their affect, and helped to explain how the bot adapts its tone to express empathy through machine learning and persona chat dialogue. The future is bright for digital assistants and developers are creating meaningful connections through learned AI behaviour.
Researchers are currently developing “empathic technology” which involves measuring peoples body temperature and carbon dioxide concentration in the breath to accurately gauge feelings and reactions. With developments like this in the pipeline, it is clear that we are only scratching the surface of what’s possible in the field of empathy AI. Watch this space.