Artificial Intelligence Chatbot for Anxiety and Depression in University Students: a Pilot Randomized Controlled Trial (Preprint)

KLOS, MARIA CAROLINA; ESCOREDO, MILAGROS; JOERIN, ANGIE; LEMOS, VIVIANA NOEMÍ; RAUWS, MICHIEL; BUNGE, EDUARDO L.

JMIR Formative Research

JMIR

Año: 2021

Background: The use of artificial intelligence based chatbots as an instrument of psychological intervention is emerging, however no studies have been reported in Latin America.Objective: This study aims to evaluate usage patterns and whether the use of a chatbot is effective for relieving depression and anxiety symptoms compared to a control group utilizing a psychoeducation book in Argentina.Methods: This was a randomized controlled trial study utilizing the chabot Tess throughout eight weeks. The initial sample consisted of 181 Argentinian college students ages 18 to 33, 87.2% female. Of those, 33 participants in the experimental condition and 30 in the control condition provided data on depressive symptoms at week eight, and 27 participants in the experimental condition and 23 in the control condition provided data on anxiety symptoms at week eight. Between and within group comparisons were analysed using Mann-Whitney U and Wilcoxon tests for depression symptoms, and Independent and Paired Samples t Tests to analyze anxiety symptoms.Results: There was no significant intergroup differences between the experimental group and the control group for depression and anxiety symptoms from baseline to week eight (P>.05). However, there were significant intragroup differences, where the experimental group showed a significant decrease in anxiety symptoms (P=.04) and no differences were observed for the controlgroup (P=.33). No significant differences were found for depressive symptoms within the groups (P>.05). The effect size of the intervention was moderate for anxiety (d=.50) and small for depression (r=.09). In regards to participants engagement after eight weeks, there was an average of 472 exchanged messages (M=472.15; SD=249.52) and a higher number of messages exchangedwith Tess was associated with positive feedback (F2,36=4.37; P=.02).Conclusions: Students engaged a considerable amount of time exchanging messages with Tess and positive feedback was associated with higher numbers of messages exchanged. The initial results show promising evidence for the use of Tess for anxiety symptoms and a lower effect on depressive symptoms in Argentinian college students. Research on chatbots is still in itsinitial stages and further research is needed.