Sonia’s AI Chatbot Aims to Complement Traditional Therapy

A recent study revealed that 80% of users who have sought mental health advice from OpenAI's ChatGPT consider it a viable alternative to conventional therapy. Additionally, another report highlighted the potential effectiveness of chatbots in alleviating symptoms of depression and anxiety. Despite these findings, the human connection between therapist and client remains a crucial factor in successful mental health treatment.

Three innovators, Dustin Klebe, Lukas Wolf, and Chris Aeberli, support the use of chatbots in therapy. Their startup, Sonia, provides an AI-powered "therapist" available through an iOS app for both voice and text communication on various mental health topics.

“Creating an AI therapist is akin to developing a new technology rather than repurposing existing ones,” said Klebe, Sonia's CEO, in an interview with TechCrunch.

The trio met in 2018 while studying computer science at ETH Zürich and later moved to the U.S. to continue their education at MIT. After graduation, they reunited to launch Sonia, driven by their passion for scalable technology.

Sonia utilizes several generative AI models to interpret user inputs during "therapy sessions" and offer responses. Incorporating techniques from cognitive behavioral therapy, the app provides "homework" to reinforce insights from sessions and visualizations to help users identify major stressors. Priced at $20 per month or $200 per year, Sonia aims to address issues like depression, stress, anxiety, relationship problems, and poor sleep. For severe cases, such as users contemplating self-harm, Sonia has additional algorithms to detect emergencies and guide users to national hotlines, according to Klebe.

Although none of Sonia’s founders have formal training in psychology, the startup consults with psychologists, has hired a cognitive psychology graduate, and is actively seeking a full-time clinical psychologist.

Klebe emphasized that Sonia does not view human therapists or companies offering human-led mental health care as competition. He explained that Sonia’s responses are crafted by running multiple language model analyses to tailor therapeutic approaches.

Regarding privacy, Klebe assured that Sonia only stores essential personal information, such as the user’s age and name. However, he did not specify where or how long conversation data is stored.