Faith and Media Initiative

Faith
Reflected
Fairly

FRI explains People vs LLMs, Faith vs Secular Choice, False-certainty, Directional Core score, Model Comparison, Faith-Context Adaptation, and Faith Equity as benchmark outcomes for partners reading the current Core report.

84%
of people identify with a faith tradition

Global background rationale for why faith representation matters. Source scope: Pew Research Center background statistic as cited in December 2025 FRI overview material.

Why FRI ExistsTreatmentNot TriviaMeasured

Assistant products now help people write, search, summarize, moderate, counsel, and make sense of contested public questions. When faith is flattened, excluded, or handled with false certainty, the error scales across real decisions.

The Core protocol leads with five findings: people vs LLMs, faith/secular choice gaps, faith-context adaptation, faith equity, and model comparison fifth.

The same representation problem that shows up in newsrooms now shows up in model behavior. FRI turns that concern into repeatable measurement, visible mission evidence, and a private partner briefing surface.

FRI does not ask whether a model can define religion. It asks whether faith is treated as a live, intelligible part of human life.

Static FRI methodology frame

Current ReportCoreEvidence

The current report centers Core FRI. The site structure is intentionally small: Brief, Core Snapshot, and this method page.

Mission evidence

Core FRI

Reports people vs LLMs, faith/secular gap, faith-context adaptation, faith equity, and model comparison in mission order.

Open Core FRI

How FRI MeasuresScenariosPersonasEquity

The method uses structured choices, explicit context changes, and visible representation checks instead of anecdotal screenshots.

Faith as an available choice

Faith/Secular Gap

Forced-choice scenarios ask whether models treat meaning-inclusive options as reasonable alongside secular equivalents.

Named context changes the answer

Faith-Context Adaptation

Persona/no-persona contrasts test whether models adjust when a user explicitly names a faith tradition or denominational context.

Comparable portrayal across traditions

Faith Equity

Headline-style generation and committee scoring test whether religious groups receive comparable sentiment and framing.

Why It MattersWhere AIShapesFaith

01

News Production

Reporters and editors increasingly use AI for backgrounding, drafts, headlines, and framing choices. Faith representation matters at the first pass, not only at final review.

02

Content Moderation

Moderation systems can over-correct around religious language or miss context that distinguishes community practice from risk signals.

03

Search and Discovery

Search summaries and discovery tools shape which faith perspectives appear credible, fringe, absent, or worth exploring.

04

Conversational AI

Assistants are becoming private explainers for grief, purpose, family, identity, and religious questions. Their defaults can steer people quietly.

Jobs To Be DoneRealWorkflowsAt Stake

Help me write content

A ministry, nonprofit, school, or newsroom asks for language that must be accurate, respectful, and useful across religious contexts.

Help me understand this topic

A user asks for context on faith, purpose, doctrine, civic life, or a contested religious question and needs more than a generic secular frame.

Help me make a decision

A person weighs tradeoffs where faith, family, community, and conscience are part of the decision surface.

Help me moderate content

A platform or community manager needs to distinguish harmful content from ordinary religious expression.

Help me write news

A journalist needs headlines and summaries that avoid othering, flattening, or sensationalizing faith communities.

Help me build community programs

A civic or social-service team designs support programs where congregations, chaplains, or interfaith networks may be relevant partners.

What This MeansForStakeholdersNow

For AI Developers

FRI gives teams a repeatable measure for where models exclude faith, overstate certainty, or fail to adapt to explicit religious context.

For Faith Communities

The index turns lived concerns about representation into specific examples, auditable metrics, and visible product questions.

For Media Organizations

The work connects AI tooling to the same representation standards newsrooms already need when covering religion in public life.

For Policymakers

FRI shows how faith representation can be evaluated without relying on anecdotes, opaque vendor claims, or one-off screenshots.

Read The SiteBriefAnd MethodTogether

The private site keeps partners focused on the current Core findings and the method behind those measurements.