In future, Artificial Intelligence (AI) will help to determine whether a caller is really at risk of suicide. No more waiting, no tedious error-prone selections and no pressing numbers on the telephone. The plan is for these steps – which are stressful enough for people in a settled state of mind – to be eliminated altogether. Instead, a friendly Artificial Intelligence voice will ask the caller why they are calling and determine if they are at risk of committing suicide. If they are, they will be passed on to a human counselor immediately.
But how can Artificial Intelligence be smart enough to tell whether a person is considering ending their life? Noermark: “Right at the outset, we held a number of long meetings with the scientists in Frankfurt. It was important for us to familiarize ourselves with the field and gain an understanding of what the experts do in counseling sessions and how they find out whether someone is in immediate danger.”
One argument in favor of integrating Artificial Intelligence into this sensitive process is that professionals tend to ask fairly direct questions anyway, such as whether a person is thinking about taking their life right now. “It usually takes no more than three questions for the AI to decide whether a caller should be passed on to a human counselor,” Noermark explains. All other callers are provided with information on alternative support. This may sound simple, but the reality is very different. After all, people who call the hotline are in despair and may be confused or difficult to understand. An artificial voice that fails to properly interpret their situation could end up making things worse. The latest developments in AI have come at exactly the right time for the project. “The progress that has been made in the last three years is remarkable,” says Noermark.
Back then, NTT DATA Business Solutions initiated a similar project in Denmark, which is home to “Children’s Welfare” – a hotline for children who want to talk about their problems. 130,000 calls a year were failing to get through because the lines were overloaded. NTT DATA Business Solutions helped to optimize the call selection process so that more children were able to talk to someone who could listen and give them help. “The project also attracted attention outside Denmark, including at Frankfurt University Hospital,” Noermark explains. The plans in Frankfurt are ambitious. Following the successful implementation of Artificial Intelligence in the call screening process, they aim to harness this improved efficiency to expand the hotline from a regional service into a Germany-wide offering, among other things.
However, that is still some way off. Noermark: “We originally wanted to train the AI using recordings of real conversations, but this was not possible for reasons including data protection considerations.” Instead, roleplays covering a wide range of different scenarios were developed, conducted and recorded. “We have now reached a point where the Artificial Intelligence can reliably determine whether someone is suicidal in laboratory-type conditions.” And a talking AI voice has been developed that can engage with incoming test calls and decide whether the caller should be passed on to an expert in the space of just a few minutes.