Exploring Customizable Interactive Tools for Therapeutic Homework Support in Mental Health Counseling

TL;DR

Therapists contend with fragmented client homework data, prompting the development of TheraTrack, a customizable, therapist-facing dashboard that centralizes heterogeneous inputs and uses GenAI to provide traceable summaries and a natural‑language chat assistant. Through a design-oriented formative study and a pilot with 14 therapists, the work demonstrates reduced cognitive load, increased access to data provenance, and adaptable workflows for private preparation versus in-session use. Key contributions include a characterization of therapists’ needs, the TheraTrack design and implementation, and empirical insights into usability, trust, and ethical considerations for clinician‑centered AI in mental health. The findings suggest that well-grounded, auditable AI can streamline between-session homework review without supplanting clinical judgment, with implications for future adoption and modality-specific customization.

Abstract

Therapeutic homework (i.e., tasks assigned by therapists for clients to complete between sessions) is essential for effective psychotherapy, yet therapists often interpret fragmented client logs, assessments, and reflections within limited preparation time. Our formative study with licensed therapists revealed three critical design requirements: support for interpreting unstructured client self-reports, customization aligned with clinical objectives, and seamless integration across multiple data sources. We then designed and developed TheraTrack, a customizable, therapist-facing tool that integrates multi-dimensional data and leverages large language models to generate traceable summaries and support natural-language queries, to streamline between-session homework tracking. Our pilot study with 14 therapists showed that TheraTrack reduced their cognitive load, enabled verification through direct navigation from AI summaries to original data entries, and was adapted differently for private analysis compared to in-session use, with dependence varying based on therapist experience and usage duration. We also discuss design implications for clinician-centered AI for mental health.

Figures (4)

• Figure 1: Overview of the three-step flow for configuring a therapist-facing tool TheraTrack. Therapist-defined need in corresponds to specific widget options in , which are then realized as interactive dashboard components in . Defining Needs: Therapists complete an onboarding survey to customize their preferences for tracking homework and client progress, such as therapists' focus areas, homework types, and assessments, setting preferences for GenAI summaries and assistance, and supplementary features including clinical information display and side functions. Choosing Widgets: Based on responses from step , therapists are guided to select relevant TheraTrack's widgets, such as chart of homework progress and health, GenAI-generated summaries and chat assistant, and message and therapy goals. Displaying Customized Dashboard: A customized dashboard will then be generated with the selected widgets, enabling therapists to choose clients, adjust widgets, and access original homework submissions, view homework trends and assessment results, interact with GenAI summaries and chat assistants, and message to their clients and modify therapy goals.
• Figure 2: Examples of therapeutic homework, such as Cognitive Behavioral Therapy (CBT) worksheets and mood trackers.
• Figure 3: Examples from the co-design sessions. A) Therapists identified and discussed key challenges they encounter when reviewing and making sense of therapeutic homework in mental health counseling, as well as the types of support they would want the system to provide for each challenge. B) Therapists generated lists of data types they considered clinically valuable, discussed what information should be included in the tool, and evaluated the feasibility of capturing or representing such data. C) Therapists reflected on broader design considerations, such as information organization, interaction flow, and conditions under which the tool would or would not be appropriate for clinical use. D) Therapists specified where and how they believed AI assistance could contribute. E) Therapists drew and annotated low-fidelity sketches.
• Figure 4: The TheraTrack main user interface showcases five example widgets. 1) Homework data overview: a bar chart visualizes the time required to complete each homework assignment, with color saturation indicating the client’s self-rated quality on a five-point scale. Two accompanying line graphs display the client’s mood ratings before and after completing each assignment, using a 1–10 scale in which 1 indicates a low mood and 10 indicates a very positive mood. 2) Health signals information, including activity levels, sleep patterns, heart rate, and mindfulness metrics. 3) GenAI-generated summary: information is organized according to priorities set by the therapist in their preferences; the system summarizes key points under ordered sections and consolidates minor points under an "Others" category. 4) Therapists can interact with a GenAI chatbot using natural language to query information about the client; responses are structured by first presenting the "relevant raw data," followed by an "AI-generated explanation." 5) Assessment results: when "Details" is selected, it expands to show all assessment items, highlighting in color any items that exceed threshold values.