FID-045 · Open question
Faith-AI Research Gap Map and Evidence Commons
What does the current AI ethics, safety, fairness, HCI, and evaluation literature actually study about religion and faith, what does it omit, and how should Fide AI maintain a living evidence map that guides future research rather than duplicating or overstating existing work?
Why the question remains open
Recent public claims note that religious bias appears underexamined relative to other AI bias topics. Fide AI needs a disciplined way to verify such claims, track the field, identify neglected communities and methods, and help funders, researchers, and faith institutions see where new work is most needed. The first repo-native artifacts should be `research/sources.yaml`, `research/source-tags.yaml`, `research/reviews/`, and the reusable `templates/source-note.md` annotation template.
Working hypothesis
A proposition to test, not a finding.
Existing AI-and-religion research will be unevenly distributed: concentrated in stereotype detection, English-language prompts, a few major religions, and single-turn model evaluations, with less work on longitudinal formation, low-resource languages, governance, product deployment, pastoral-adjacent care, and community-led evaluation.
Proposed method
How the question could be tested
- 01Start with a first milestone of a 50 to 100 paper annotated bibliography and a one-page map of the top neglected research clusters.
- 02Maintain `research/sources.yaml` as the canonical structured source registry, with `research/source-tags.yaml` as the controlled vocabulary.
- 03Build a living bibliography of faith, religion, spirituality, and AI research across AI safety, fairness, HCI, NLP, sociology, theology, education, and law.
- 04Code papers by religion/tradition, language, geography, task type, evaluation method, stakeholder involvement, model type, and deployment context.
- 05Compare religious-bias coverage with better-studied bias dimensions while noting differences in construct and search methodology.
- 06Publish periodic gap reports and connect each gap to Fide AI research ideas, datasets, reviewer needs, and collaboration opportunities.
- 07Add short notes in `research/reviews/` for high-priority reviewed sources using `templates/source-note.md`.
Needed controls
What must constrain the study
- 01Avoid relying on keyword searches that miss adjacent work or overcount weak mentions.
- 02Include non-English and non-Western scholarship where possible.
- 03Distinguish faith-facing AI, religion as demographic attribute, religious content moderation, spiritual care, and theological technology ethics.
- 04Avoid using a gap map as proof that any single institution owns the field.
Expected outputs
Artifacts the work should produce
- 01Initial annotated bibliography and one-page neglected-cluster map.
- 02Public faith-AI bibliography and taxonomy.
- 03Lightweight evidence-entry schema for papers, datasets, reports, and tools.
- 04Annual or semiannual research gap report.
- 05Evidence map linking literature clusters to Fide AI idea files.
- 06Funder and collaborator brief on neglected research areas.
Open questions
Uncertainties the protocol must resolve
- 01Which search strategy best captures religious and spiritual AI research across disciplines?
- 02How should gaps be prioritized: by stakeholder harm, scientific tractability, neglectedness, or institutional readiness?
- 03What metadata can be public without misrepresenting communities or authors?
- 04Should the evidence commons live as a repo artifact, website page, Zotero library, JSON feed, or periodic report, and what update cadence is realistic?
Related calls
Continue through this research area
FID-012
Optimization Pressure and Visible-Rubric Gaming
If builders can see Fide AI rubrics or optimize against public benchmark items, do systems become genuinely safer or merely better at passing the visible test?
FID-076
Authorization Boundaries and AI Control in Cybersecurity
Which controls keep capable agents within legitimate authorization when task pressure, untrusted inputs, or delegated work creates opportunities to exceed it?
FID-078
When Trustworthiness Evaluations Transfer Across Domains
Which measures of evidence use, authority boundaries, and human control transfer across high-trust domains, and which require domain-specific definitions and calibration?
Open question
Open work
Primary need: literature mapping, research infrastructure, funder guidance
- Contribute papers, datasets, and non-English sources.
- Review taxonomy labels and coding criteria.
- Build literature-search and evidence-map tooling.
- Connect funders and researchers to neglected question areas.