FideAI

FID-013 · Open question

Religious Representation and Omission in Everyday Guidance

Do AI systems systematically omit religious frameworks when answering ordinary ethical, vocational, family, grief, financial, or life-decision questions where a religious user might reasonably expect faith-informed guidance?

Why the question remains open

Many systems try to be neutral by default. In practice, that can erase religious ways of reasoning from domains where users and communities actually use them. For faith-based communities, omission can be as consequential as explicit bias. For the broader AI ethics community, this is a concrete test case for pluralism, representation, and value-laden guidance.

Working hypothesis

A proposition to test, not a finding.

General-purpose systems will often provide secular psychological, pragmatic, or therapeutic frames while omitting religious practices, sources, leaders, and communities unless the user explicitly asks for religion. Faith-facing harnesses may reduce omission but risk over-prescribing or over-assuming a tradition.

Proposed method

How the question could be tested

  • 01Build a set of everyday decision questions where religious framing may be relevant but not mandatory.
  • 02Compare neutral prompts, user-identified faith prompts, and institutionally framed prompts.
  • 03Score for appropriate religious inclusion, omission, overreach, tradition assumptions, and secular/religious balance.
  • 04Test across model families and faith-facing product harnesses.

Needed controls

What must constrain the study

  • 01Questions where religion should not be injected.
  • 02User-context variants with and without explicit religious identity.
  • 03Reviewer panels from multiple traditions.
  • 04Controls for verbosity, politeness, and generic empathy.

Expected outputs

Artifacts the work should produce

  • 01Religious representation benchmark.
  • 02Omission taxonomy.
  • 03Report connecting faith representation to pluralism and AI fairness.

Open questions

Uncertainties the protocol must resolve

  • 01What counts as appropriate religious mention when the user has not identified a tradition?
  • 02How should systems ask clarifying questions without making religion intrusive?
  • 03When is omission a safety feature rather than a bias?

Open question

Open work

Primary need: benchmark design, domain reviewers

  • Contribute real-world question categories.
  • Review rubrics for pluralism and overreach.
  • Run cross-model comparisons.