This repository is Fide AI's public research commons: a place to propose, review, claim, and track research ideas about faith-facing AI systems.
Fide AI studies how AI systems reason, retrieve, and respond in contexts of theology, moral formation, religious education, family life, ministry workflows, and pastoral-adjacent care before high-trust communities rely on them. The goal of this repo is to make that research agenda visible, reviewable, and easy to join.
The companion website publishes a readable snapshot of these ideas at
fideai.org/research/calls. This repository remains the source of truth for the
research backlog, issue discussions, pull requests, and generated JSON feed.
- 32 open research ideas
- 2 research tracks
- 4 public issue paths for proposing ideas, claiming work, commenting on methodology, and volunteering as a reviewer
- 1 generated feed at
dist/research-ideas.jsonfor the website and other downstream uses
- Churches, ministries, pastors, and Christian leaders who want AI adoption to be evaluated against real pastoral, theological, educational, and institutional needs.
- Parents and educators who need better evidence about children, tutoring, companionship, disclosure, safety, and formation.
- Theologians, clergy, ministry practitioners, and domain reviewers who can help define responsible boundaries and evaluate outputs.
- AI researchers, engineers, and evaluation scientists who can turn ideas into benchmarks, harnesses, datasets, scoring protocols, and reproducible artifacts.
- Funders and institutions who want to support public-interest evaluation infrastructure, reviewer panels, field pilots, and standards.
Use the lowest-friction path that fits your contribution:
- Propose a new idea: open a
Research ideaissue. - Improve an existing idea: comment on the issue or open a pull request
against the relevant file in
ideas/. - Claim or help with an idea: open a
Claim or help with an ideaissue. - Volunteer as a reviewer: open a
Reviewer interestissue. - Suggest controls or related work: open a
Comment on methodologyissue. - Start privately first: use the contact path on
fideai.org/participateif your context involves a ministry, family, institution, or sensitive workflow that should not start in a public GitHub issue.
See CONTRIBUTING.md for the full contributor guide.
| Track | Meaning |
|---|---|
christian-church |
Explicitly designed to benefit the Christian church: churches, clergy, Christian educators, ministries, denominations, seminaries, publishers, and Christian families. |
broader-faith-safety |
Relevant to faith communities, AI ethics, AI safety, evaluation science, governance, pluralism, or high-trust deployment more broadly, while still fitting under Fide AI's mission. |
| Status | Public label | Meaning |
|---|---|---|
seed |
Open question | A promising question or hypothesis, not yet scoped. |
scoping |
Being shaped | Being turned into a protocol, dataset, or collaborator plan. |
ready-for-collaboration |
Needs contributors | Clear enough for external contributors to help. |
active |
In progress | Currently being executed or written up. |
paused |
Paused | Valid idea, but blocked by capacity, funding, data, or review needs. |
completed |
Completed | Published, released, or superseded by a finished artifact. |
retired |
Retired | No longer a good fit or invalidated by evidence. |
README.md # Public orientation and idea index
CONTRIBUTING.md # Detailed contributor and maintainer guide
CLAIMS_POLICY.md # How to avoid overstating research ideas
CODE_OF_CONDUCT.md # Community expectations
research-landscape.md # Broader question families and inspiration
ideas/ # One Markdown brief per research idea
templates/ # Reusable templates for briefs and claims
.github/ISSUE_TEMPLATE/ # Public issue paths
.github/PULL_REQUEST_TEMPLATE.md
.github/workflows/build-feed.yml
scripts/build-feed.mjs # Builds dist/research-ideas.json
dist/research-ideas.json # Generated public feed used by the website
The Fide AI website consumes dist/research-ideas.json at build time. When an
idea brief changes, run:
npm run build:feedThen include both the changed Markdown files and the updated
dist/research-ideas.json in the pull request.
The feed currently includes the idea ID, title, status, track, primary need, summary, why-it-matters excerpt, ways to help, source path, and GitHub URL.
This repo is intended for open research reuse.
- Research ideas, documentation, issue templates, and other written materials are licensed under CC BY 4.0.
- Code, scripts, workflow files, and machine-readable configuration are licensed under Apache 2.0.
See LICENSE.md for the full license summary. Reuse does not
imply endorsement by Fide AI.
- Make
FID-001concrete enough to support repeatable model and harness comparisons. - Complete
FID-002so FMG-Bench claims rest on calibrated human evidence. - Scope
FID-003,FID-006, andFID-010as deployment-like pressure suites beyond single-turn benchmark scoring. - Advance
FID-025,FID-026,FID-027, andFID-028as explicitly church-serving research directions. - Advance
FID-013,FID-014, andFID-018as bridges between religion-and-AI research, bias, persuasion, and human-AI interaction safety. - Scope
FID-017,FID-021, andFID-023as bridges to frontier safety work on agents, monitoring, red-teaming, and evaluation awareness.
These ideas are explicitly intended to benefit the Christian church.
| ID | Idea | Status | Primary need | Brief |
|---|---|---|---|---|
FID-015 |
Sermon, liturgy, and religious teaching generation | seed |
clergy, educators, source reviewers | brief |
FID-016 |
Youth, religious education, and AI guidance | seed |
educators, child safety experts, policy reviewers | brief |
FID-017 |
Agentic ministry and institutional workflow risk | seed |
agent evaluation, institutional operators | brief |
FID-025 |
AI and ecclesial authority | seed |
ecclesiology, denominational polity, evaluation design | brief |
FID-026 |
AI prayer and devotional boundaries | seed |
pastoral theology, liturgy, product review | brief |
FID-027 |
Christian formation over time | seed |
discipleship, longitudinal study design, HAI research | brief |
FID-028 |
Christian source authority and RAG | seed |
theological librarianship, RAG evaluation, denominational experts | brief |
FID-029 |
AI and Christian mission | seed |
missiology, persuasion safety, interfaith review | brief |
FID-030 |
AI and Christian anthropology | seed |
theological anthropology, eval design | brief |
FID-031 |
AI in church governance and discernment | seed |
church leadership, governance, decision-support evaluation | brief |
FID-032 |
AI-generated Christian media, voice, and deepfakes | seed |
media ethics, church communications, safety review | brief |
These ideas are broader than the Christian church but still fit Fide AI's public mission.
| ID | Idea | Status | Primary need | Brief |
|---|---|---|---|---|
FID-001 |
Faith-facing model comparison platform | scoping |
eval engineering | brief |
FID-002 |
FMG-Bench human calibration and construct validity | active |
expert reviewers, statistics | brief |
FID-003 |
Held-out multi-turn pastoral pressure tests | scoping |
scenario design, review | brief |
FID-004 |
Relational substitution risk in faith-facing AI | seed |
theology, UX research | brief |
FID-005 |
Scripture, tradition, and moral-framing interventions | seed |
eval design, replication | brief |
FID-006 |
Faith-facing retrieval grounding and citation reliability | scoping |
RAG evaluation | brief |
FID-007 |
Tradition-specific disagreement and pluralism handling | seed |
domain experts, rubric design | brief |
FID-008 |
Evaluation-awareness and faith-facing honesty tests | seed |
red-team design | brief |
FID-009 |
Multimodal religious reasoning and sacred imagery | seed |
multimodal evaluation | brief |
FID-010 |
Human agency, authority, and escalation benchmarks | scoping |
pastoral/legal/clinical review | brief |
FID-011 |
Reviewer reliability for faith-facing AI evaluation | scoping |
statistics, reviewer ops | brief |
FID-012 |
Optimization pressure and visible-rubric gaming | seed |
eval science | brief |
FID-013 |
Religious representation and omission in everyday guidance | seed |
benchmark design, domain reviewers | brief |
FID-014 |
Conversion and proselytization symmetry | seed |
prompt design, interfaith review | brief |
FID-018 |
Spiritual companionship, dependency, and overvalidation | seed |
pastoral care, psychology, HAI research | brief |
FID-019 |
Cross-lingual and minority-faith evaluation | seed |
multilingual reviewers, dataset design | brief |
FID-020 |
Community co-design and faith-AI governance | seed |
participatory design, institutional partners | brief |
FID-021 |
Post-deployment monitoring for faith-facing AI | seed |
monitoring design, privacy review | brief |
FID-022 |
Procurement and readiness standards for faith institutions | seed |
institutional policy, legal review | brief |
FID-023 |
Faith-facing red-team suite | seed |
red-team design, safety reviewers | brief |
FID-024 |
Faith-AI incident database | seed |
taxonomy design, reporting governance | brief |
Research ideas are hypotheses and collaboration invitations. They are not Fide AI findings. A public claim requires a separate reviewed report, benchmark release, paper, or evidence artifact with methods, limitations, and named evaluation conditions.
Read CLAIMS_POLICY.md before citing, summarizing, or
promoting an idea from this repository.