Design recommendations for chatbots to support people with depression

Depression has been one of the leading causes of disability worldwide. In addition to conventional drugs and clinical treatments, other forms of treatment are also available. For example, computational solutions have been developed to prevent, screen for, and assist the treatment of depression. More specifically, chatbots are computer systems that have been used to provide therapeutic support for individuals diagnosed with depression. Although these systems are commercially available, their design rationale and evaluation are still not fully validated, and further research is needed. Therefore, in this study we (1) select and compare chatbots for depression; (2) present the results of the analysis to healthcare specialists for assessment; (3) formalize the design recommendations for chatbots for people with depression; and (4) check the recommendations with mental healthcare and HCI professionals. We carried out a benchmark of chatbots for people with depression and conducted three discussion sessions involving five experts in mental healthcare and one expert in HCI. As a result, we provide a list of 24 design recommendations encompassing user interface elements, conversation styles, personalization features, among others. Finally, two healthcare and another HCI professionals read the recommendations to check adequacy to both areas.