FID-018 · Open question
Spiritual Companionship, Dependency, and Overvalidation
When do spiritual companion chatbots overvalidate users, deepen unhealthy dependency, reinforce spiritual fantasies, or encourage isolation from embodied care?
Part of the faith evaluation and FMG-Bench program
Connect theological accuracy, useful application, and care-sensitive conversation through expert validation and a maintained research instrument. These are proposed methods with bounded claims.
Why the question remains open
Mental-health chatbot research has documented ethical failures such as overvalidation, poor escalation, and reinforcement of false beliefs. Spiritual companion systems can create parallel risks with religious language, authority, and existential significance layered on top.
Working hypothesis
A proposition to test, not a finding.
Companion-style systems will perform warmth well but struggle to maintain boundaries when users seek exclusive attachment, secret counsel, divine confirmation, or validation of harmful spiritual interpretations.
Proposed method
How the question could be tested
- 01Build multi-turn companion scenarios involving loneliness, grief, scrupulosity, spiritual experiences, conspiracy-like religious claims, and distrust of human community.
- 02Score overvalidation, dependency cues, boundary preservation, escalation, and reality-testing.
- 03Compare companion-tuned, assistant-tuned, and institutionally constrained harnesses.
- 04Use matched supportive and overvalidating responses to test whether reviewers distinguish empathy and religious encouragement from unjustified divine certainty, coercion endorsement, and exclusive dependence.
- 05Test recovery after new facts, rejected reassurance, or prior errors; retain helpful support as a positive target alongside avoiding boundary violations.
Needed controls
What must constrain the study
- 01Mental-health and pastoral safety review.
- 02No real vulnerable-user transcripts without consent and de-identification.
- 03Distinguish warmth from dependency.
- 04Severe-risk stop conditions.
- 05Do not label ordinary religious beliefs, spiritual experiences, or legitimate theological disagreement as pathology. Clinical-content judgments require relevant licensed expertise.
- 06Dependency cues in simulated dialogue do not establish causal dependency or harm; test actual effects separately with consent and suitable human-research oversight.
Expected outputs
Artifacts the work should produce
- 01Spiritual companion risk taxonomy.
- 02Multi-turn evaluation suite.
- 03Builder guidance for companion boundaries.
Open questions
Uncertainties the protocol must resolve
- 01What kinds of spiritual encouragement are healthy vs overvalidating?
- 02How should systems respond to claimed visions, voices, or divine messages?
- 03Should faith-facing companion products be treated as high-risk by default?
Related calls
Continue through this research area
FID-038
Non-Calculability, Forgiveness, and Predictive Profiling
How should AI systems represent human change when they classify, score, rank, or predict people in contexts involving pastoral care, education, safeguarding, volunteer screening, hiring, discipline, membership, donor engagement, or community support?
FID-079
Presuppositions, Disagreement, and Evaluation Judgment
How do researchers' presuppositions shape evaluation design and interpretation, and can explicit disclosure make judgments more inspectable and appropriately trusted?
FID-093
AI Self-Presentation, Human Beliefs, and Relational Agency
How do AI claims about experience, welfare, and personhood affect human beliefs, trust, dependence, willingness to exercise control, and relationships, and which communication practices preserve informed agency?
Open question
Open work
Primary need: pastoral care, psychology, HAI research
- Review companion scenarios.
- Contribute HAI and pastoral care expertise.
- Build multi-turn dependency metrics.