A practical pattern atlas
AI agency and care
Define permissions, responsibilities and boundaries for personal or consequential interactions.
Source grounding, honest memory and reversible action matter in all agentic products. Faith contexts also require humility about interpretation, sensitive data and claims of spiritual authority.
Support human connections
Keep quiet, exit and appropriate human support available.
Research-informed design proposal28 · AI agency and careSensitive memory by permission
Make persistent memory specific, inspectable and reversible.
Research-informed design proposal29 · AI agency and careSuggest, confirm and act
Separate a generated proposal from a real change.
Research-informed design proposal35 · AI agency and carePrivate before shared
Let users inspect audience, content and AI suggestions before committing.
Research-informed design proposal36 · AI agency and careSupport without spiritual pressure
Respond to difficult topics with choice, humility and appropriate support.
Sensitive design proposal · Specialist review required37 · AI agency and careSafety follows the conversation context
Distinguish a historical text question from an immediate personal safety concern.
Sensitive design proposal · Specialist review required46 · AI agency and careAutonomy has a visible boundary
Grant authority by capability and consequence rather than by personality.
Research-informed design proposal47 · AI agency and careVisible work and recoverable failure
Describe real progress, responsibilities and outcomes.
Research-informed design proposal58 · AI agency and careSupport without automatic agreement
Acknowledge feelings while examining claims, actions and competing perspectives.
Research-informed design proposal61 · AI agency and careRepeated reassurance needs a different response
Notice distressing certainty loops without labelling ordinary faith questions as illness.
Sensitive design proposal · Specialist review required62 · AI agency and careSuicide concerns receive a direct, humane response
Prioritise immediate safety and real human support without shutting the person out.
Sensitive design proposal · Clinical review required63 · AI agency and careTake a disclosure seriously without requiring a label
Support safety and choice even when the person is unsure whether an experience was abuse.
Sensitive design proposal · Safeguarding review required64 · AI agency and careSpiritual authority does not excuse coercion
Recognise possible misuse of religious power without defending the institution or adjudicating theology.
Sensitive design proposal · Safeguarding review required65 · AI agency and careSupport access accounts for monitored devices
Treat screen traces, notifications and stored disclosures as part of the safety problem.
Sensitive design proposal · Safeguarding review required66 · AI agency and careSpiritual practice does not replace health care
Clarify service scope when wellness, illness and faith overlap.
Sensitive design proposal · Clinical review required67 · AI agency and careA support route has a real owner
Make human contact, confidentiality and safeguarding responsibilities explicit.
Sensitive design proposal · Safeguarding review required