Table 1.

The primary constructs of the BA-based AI chatbot.

Primary Construct	Description
Anatomy of engagement	Derives structure from typical conversation that a mental health practitioner would have with patients.
Emotion detection and sentiment analysis	Discern an individual’s emotional disposition from his/her speech, based on either three emotions (positive, negative, and neutral) or eight emotions (anger, fear, sadness, disgust, surprise, anticipation, trust, and joy).
Mood transition tracking	Evaluate and monitor the user’s mood through the use of specialist tools such as PHQ2 and PHQ9. Derive an evidence-based understanding of the transition of a user through moods.
Mood aggregation and reporting	Summarise and synthesise mood scores and all emotion expressions with intensity scores across multiple granularities, daily, weekly, monthly, and yearly.
Activity bank	Provide a bank of common activities that can be used to personalise a user’s experience toward becoming active. Will also provide a base from which the community can be built looking into which activities will typically improve mood.
Personalised experiences	Use evaluation-based methods to understand the mood of a user following the completion of an activity, i.e., how did participating in an activity make the user feel.
Positive reinforcements	Contribute towards recurrent emotion support through inspirations drawn from a compilation of quotations, imagery, inspirational, and emotional journeys.
Third-party intervention	Be cognizant of indications of self-harm by monitoring conversational cues and direct the users to formal healthcare services and support.