Chaplains' Reflections on the Design and Usage of AI for Conversational Care

Paper Title

Chaplains' Reflections on the Design and Usage of AI for Conversational Care

Publication Info

• Topic area: Exploring chaplains' perspectives on AI chatbots for non-clinical well-being support.
• Keywords: AI chatbots, pastoral care, well-being, conversational AI, attunement, chaplains, emotional support, listening, connecting, human presence.

Background and Problem

• Problem / challenge: Current conversational AI systems are primarily designed with clinical expertise, focusing on diagnosis and intervention, and fail to address the non-clinical emotional support needs that chaplains provide, such as listening, connecting, and carrying emotional burdens.
• Significance: Understanding chaplains’ perspectives can inform the design of AI chatbots to better support well-being in non-clinical contexts, addressing gaps in emotional and relational care.
• Motivation and related work: Prior research has explored parasocial support and mental health applications of AI chatbots, but little attention has been given to how chaplains’ expertise in pastoral care can inform chatbot design. This study addresses this gap by examining chaplains’ views on the limitations and potential of AI chatbots.

Solution

• Proposed approach: Engaging chaplains in designing and reflecting on AI chatbots using the GPT Builder interface to explore their perspectives on conversational care and chatbot limitations.
• Novelty:
  1. Involving chaplains as domain experts in the design of AI chatbots for non-clinical well-being.
  2. Identifying four key themes—Listening, Connecting, Carrying, and Wanting—that highlight gaps in chatbot capabilities compared to human pastoral care.
  3. Introducing the concept of attunement as a framework for understanding and improving chatbot design for care contexts.
• Procedure and key techniques:
  • Recruited 18 Nordic university chaplains to design and test chatbots for fictional student profiles using GPT Builder.
  • Conducted pre- and post-intervention interviews to elicit chaplains’ reflections on chatbot design and their professional practices.
  • Analyzed transcripts to identify themes and insights into chaplains’ perspectives on AI chatbots.

Results

• Concrete findings:
  • Chaplains rated generative AI acceptance moderately (median overall score: 62.5/100).
  • Four themes emerged:
    1. Listening: AI chatbots fail to replicate the nuanced, agenda-free listening chaplains provide.
    2. Connecting: Chatbots lack the emotional and physical presence necessary for genuine connection.
    3. Carrying: Chatbots cannot take responsibility for or emotionally hold users’ concerns.
    4. Wanting: Chatbots exhibit eagerness (e.g., rapid responses, verbose outputs) that contrasts with chaplains’ reflective, goal-less approach.
• Advantage over baselines: Highlights unique insights from chaplains, contrasting with prior studies that focus on clinical or end-user perspectives.
• Experiments / evaluation:
  • Chaplains created four chatbots each, including one general-purpose and three tailored to fictional student profiles (Maya, Leo, Samira).
  • Reflections captured through interviews and analyzed for thematic insights.
• Limitations and future work:
  • Limited generalizability due to Nordic context and small sample size.
  • Chaplains had minimal prior experience with AI chatbots, potentially influencing their designs.
  • Future work should involve end-users and explore interactive prototypes informed by chaplains’ insights.

Summary

This study explored how chaplains perceive and design AI chatbots for non-clinical well-being support, identifying key limitations in chatbots’ ability to replicate human presence and care. Four themes—Listening, Connecting, Carrying, and Wanting—highlighted gaps in chatbot design compared to pastoral care practices. The concept of attunement was introduced as a framework for improving chatbot interactions. These findings offer valuable insights for designing AI systems that better align with human emotional and relational needs in non-clinical care contexts. Future work should integrate chaplains’ expertise with end-user perspectives to develop more attuned conversational AI.