When Your Therapist Is an Algorithm: Understanding the Role of AI in Mental Health Mobile Applications

Paper Title

When Your Therapist Is an Algorithm: Understanding the Role of AI in Mental Health Mobile Applications

Publication Info

• Topic area: The integration and user experience of AI in mental health mobile applications.
• Keywords: AI in mental health, digital mental health, user experience, AI roles, chatbots, therapeutic support, trust in AI, ethical AI, personalization, app design.

Background and Problem

• Problem / challenge: Despite the increasing integration of AI in mental health apps, there is limited understanding of how AI is implemented, the roles it plays, and how users perceive and experience these features. Prior research has not systematically analyzed AI's functional roles or user feedback in this domain.
• Significance: Mental health apps offer scalable, accessible support, but their effectiveness hinges on user engagement, trust, and ethical design. Understanding AI's role and user perceptions is critical for improving these technologies and addressing unmet mental health needs.
• Motivation and related work: Previous studies have explored the usability, trustworthiness, and content quality of mental health apps but have not focused on the specific roles AI plays or the tensions users experience. This paper addresses these gaps by mapping AI roles and analyzing user feedback to inform better design strategies.

Solution

• Proposed approach: A two-stage analysis combining a systematic review of AI-enabled mental health apps and a large-scale analysis of user reviews to map AI roles and understand user perceptions.
• Novelty:
  1. Identification and classification of 12 distinct AI roles in mental health apps.
  2. Large-scale sentiment and thematic analysis of 996 user reviews to uncover user experiences and tensions.
  3. Design recommendations addressing core tensions in AI integration: replacement vs. supplementation, trust, and augmentation.
• Procedure and key techniques:
  1. Systematic review of 244 AI-enabled mental health apps from the Apple App Store to classify AI roles and interface types.
  2. Sentiment analysis of 996 user reviews to quantify positive, negative, and mixed perceptions.
  3. Reflexive Thematic Analysis (RTA) to identify recurring tensions and values underlying user feedback.
  4. Development of design recommendations based on findings.

Results

• Concrete findings:
  • Identified 12 AI roles (e.g., personal coach, tracker, companion, summarizer, therapeutic support) and 4 interface types (invisible AI, chatbot, voice interface, embodied agent).
  • Sentiment analysis revealed that functional roles (e.g., tracker, teacher) received more positive feedback, while relational roles (e.g., companion, therapeutic support) elicited mixed or negative sentiment.
  • Three recurring tensions emerged: AI replacing vs. supplementing human care, trust in AI's reliability and ethics, and the effectiveness of AI in augmenting self-care.
• Advantage over baselines: The study provides a comprehensive mapping of AI roles and user experiences, offering actionable insights for designing empathetic and trustworthy mental health apps.
• Experiments / evaluation:
  • Dataset: 244 apps and 996 user reviews collected from the Apple App Store.
  • Methods: Sentiment analysis using GPT-4o and thematic analysis to interpret user feedback.
  • Metrics: Sentiment distribution, thematic patterns, and role-specific user experiences.
• Limitations and future work:
  • Limited to apps explicitly mentioning AI and reviews in English.
  • Focused on the Apple App Store; future work could include other platforms and multilingual datasets.
  • Sentiment analysis relied on GPT-4o, which may introduce classification variability; future studies could compare multiple analysis methods.

Summary

This study systematically analyzed the roles of AI in mental health apps and user perceptions of these features. It identified 12 AI roles and 4 interface types, revealing tensions around AI replacing human care, trust in AI, and its ability to augment self-care. Sentiment and thematic analysis of 996 user reviews highlighted the importance of designing AI as a supportive, trustworthy, and adaptive partner rather than a substitute for human care. The findings contribute to human-centered design strategies for ethical and effective AI in mental health technologies.