Sprechstunde mit dem Roboter

"How can I help you?" That's how many conversations start. Increasingly, however, this question is not being asked by a person, but by a machine. Chatbots are the name of the helpers that 63 percent of Swiss people have already talked to, according to a recent study. They greet us on websites, in apps or social media - in other words, wherever we interact with the digital world. Online shops have been using them for a long time, as have vacation destinations, insurance companies and administrations. And in the coronavirus year, a chatbot called Replika, with which isolated people can enter into a romantic relationship, even made a name for itself.

There are also many conceivable applications for chatbots in the healthcare sector. They can - to a certain extent - replace the conversation with the therapist or doctor: Using a structured dialog, they ask standard questions, similar to those asked in a consultation. As a health coach, a chatbot can help with questions about nutrition and sleeping habits, for example, or how to deal with a newly diagnosed illness. He can do what we do when we talk to each other: impart knowledge. Today, many people obtain information on health topics on the Internet. On the go, at work or on the sofa in the evening. And this is the big advantage of chatbots: their availability. Dr. Chatbot always has consultation hours. The digital helper also performs administrative tasks. A chatbot captures boring forms playfully by asking a few questions.

In its simplest form, the dialog is achieved by an algorithm that selects the answer that matches the user's question from a list of ready-made answers. If the chatbot gets stuck, a real person can continue the conversation. Chatbots based on artificial intelligence (AI) go further. They analyze the text input and provide their own answers based on the data collected. Over time, such a chatbot "learns" to respond more and more specifically to the language of its counterpart and thus become a more "empathetic" conversation partner. One example of this is the Replika app mentioned above. Speaking of conversation: The majority of chatbots today are text-based, but there are also voice-based variants. Siri, Alexa and co. are ultimately also chatbots.

Like chatbots themselves, research into their benefits in the healthcare sector is still in its infancy. A team from ETH Zurich and the University of St. Gallen has therefore developed MobileCoach - an open source platform that companies and researchers around the world can use to develop their own chatbots. "Replicability is extremely important in science," says Dr. Tobias Kowatsch, who initiated and co-developed MobileCoach. "Anyone can basically understand how the platform works."

MobileCoach is also rule-based, but can remember previous interactions and access sensor data from smartwatches etc. According to Kowatsch, more research is needed for chatbots to become established in medicine. "Once reliable data on effectiveness is available, acceptance will increase among both users and healthcare professionals."

Someone who uses MobileCoach is Sandra Hauser-Ulrich, who conducts research into digital health applications in coaching at the Zurich University of Applied Sciences ZHAW. In one study, she used a chatbot to help people with chronic pain. The results astonished them: "Although the participants knew that they were communicating with a machine , they built up a relationship."

However, there are limits: "A rule-based chatbot cannot react to spontaneous thoughts. That's why it will never replace real therapy with a human." AI chatbots, on the other hand, are not yet ready to maintain a longer conversation - the diversity of human language with its immense amount of terms and idioms is too great. "AI fails at chatting," says Hauser-Ulrich. So if you want to have a chat with the doctor, you can't avoid a real consultation. At least for the time being.