Chatbots Are Getting Lessons In Empathy From A Machine

Koko emulates human-like responses to make AI interactions more fluid and productive

Chatbots are not known for their sensitivity or their human-like responses: ask Siri or Alexa what to do if you’re stressed or depressed and they won’t lend much of a sympathetic ear. Koko, however, is a new service that could make bot interactions more beneficial to those who are suffering.

First crafted in 2014 by MIT’s Media Lab as a program that connected people with mental strife to others in similar situations, Koko is now creating new algorithms in order to better help its user base. Currently, it is being used to funnel requests to the most appropriate matches, though it often has conversations with the people until someone on the other side is available. The more conversations it monitors, the more the AI learned to replicate the specific language and tonal patterns that people use when discussing sensitive subjects.

After finishing their $2.5 million Series A round of financing, the chatbot has been extended to work with text chat services like Telegram and Facebook Messenger. The goal is to eventually integrate it into audio chatbots like iOS’s Siri, Android’s Cortana, and Amazon’s Alexa. Co-creator Fraser Kelton tells co. Design that the ultimate end game is to make a human-computer conversation seamless:

“We’re working toward providing empathy as a service to any voice or messaging platform. We think that’s a critical user experience for a world in which you’re conversing with computers.”

Koko is about six months away from launching an app for Amazon, making Alexa one of the first audio AIs to have more empathetic responses to anxiety, depression, stress, and other heightened emotional and mental states.