Comprehensive Summary
This study, presented by Zhao et al., aimed to test the effectiveness of an AI conversational agent called Douyin companion bot at reducing young people’s negative emotions. 865 participants aged 18-25 were recruited from the app Douyin and separated into the AI conversation intervention group and the waiting group. The AI conversation intervention group was asked to have five minute conversations with the chatbot about their feelings every day for 28 days and was given three questionnaires throughout the intervention. These questionnaires measured depression, anxiety, and general mood. The waiting group waited for the first four weeks, throughout which they were given the same questionnaires at the same time as the intervention group. They then started the conversational intervention but received no further questionnaires. After fourteen days of AI conversation intervention, there were no differences from baseline in any moods for the intervention and waiting group. After 28 days of intervention, the intervention group showed a significant decrease in depression, anxiety and negative mood, and the waiting group showed a significant decrease in positive mood. Results of ANOVA showed that, in the intervention group, intervention reduced depression after two weeks and reduced both depression and anxiety at four weeks. This suggests that the intervention may have more immediate effects on depression than anxiety. The results indicate that the Douyin companion bot is successful at reducing negative emotions in young adults and that the training that was used to develop it is effective.
Outcomes and Implications
Depression and anxiety are prevalent among young adults, but many young adults do not seek treatment due to factors such as shame and lack of education about mental health. Mobile-based interventions such as the Douyin companion bot are more accessible than traditional interventions and reduce the effects shame and stigma may have on reaching out. They can appear as pop-ups when people search for topics such as depression and anxiety and offer immediate support. However, this study has multiple limitations: its duration was short-term, there was a significant sex difference between intervention and waiting groups, there was structured attrition of subjects, participants were aware that the study was psychology-related, this study only showed what happens when the AI chatbot is used as intended, which may not happen in real life, and adolescents were not tested despite being a potential target for this intervention. Therefore, the chatbot is limited in its applicability and findings cannot be generalized.