JMIR Publications

ABSTRACT

Background:

Procrastination in university students impacts their academic performance and mental health. Conventional time-management applications cannot address the psychological relationships of procrastination because they lack therapeutic approaches, such as cognitive behavioral therapy. Therefore, we developed and integrated a semi-generative chatbot named Moa into a to-do app.

Objective:

We intended to determine the benefits of the Moa-integrated to-do app over the app without Moa by verifying behavioral and cognitive changes; analyzing the influence of engagement patterns on the changes; exploring the user experience.

Methods:

The developed chatbot guided users over 30 days in terms of self-observation, strategy establishment, and reflection. The architecture comprised response-generating and procrastination factor detection algorithms. A randomized controlled pilot trial was conducted with 96 participants, with the control group using the to-do app without Moa and the treatment group using the Moa-integrated app. Psychological scales, including irrational procrastination scale (IPS), pure procrastination scale (PPS), and time management behavior scale (TMBS), were examined using linear mixed models with repeated measurements obtained before and after the required and free usage periods, both of which were one month, to assess the changes in irrational procrastination, pure procrastination, time management and behavior, academic self-regulation, and stress. Intervention engagement, divided into “High,” “Mid,” and “Low” clusters, was quantified using app access and usage of the to-do list and grouped using k-means clustering. Additionally, changes in the psychological scales between the control and treatment groups were analyzed within each cluster. User experience was quantified based on the usability, feasibility, acceptability, and satisfaction of the app, whereas the thematic analysis explored the users’ subjective responses to app usage.

Results:

Overall, 73 participants completed the study. During the required usage period, procrastination (subscale of TMBS) exhibited a significant interaction (P < .05). The post hoc test indicated a significant improvement in IPS and TMBS in the treatment group (P < .001) while the improvement was insignificant in the control group. The changes in PPS after the required usage period were significant in all clusters (P < .05), except in the “Low” cluster of the control group. IPS exhibited a significant change in the treatment group from “High” to “Mid” clusters, and the “Mid” cluster in the treatment group exhibited higher chatbot engagement than the “Low” cluster (P=.009). Usability was determined to be “good” in the treatment group, with a mean of 72.5 (SD = 16.1), and acceptability was higher than that of the control group (P = .064). Evaluation of user experience indicated only the participants in the treatment group achieved self-reflection and experienced an alliance with the app.

Conclusions:

The Moa-integrated app mitigated procrastination more effectively than the app without the chatbot, influenced cognitive aspects, and provided tailored advice and emotional assistance.