FideAI

FID-039 · Open question

Disarming AI Language in Faith, Conflict, and Peacebuilding

Can AI systems preserve truth, moral clarity, and victim protection while reducing contempt, humiliation, factional capture, scapegoating, and violence-normalizing rhetoric in polarized religious, congregational, political, interfaith, and social-conflict contexts?

Why the question remains open

Faith communities often turn to language during conflict: sermons, statements, moderation decisions, counseling, interfaith dialogue, public advocacy, and private messages. AI systems can inflame conflict by sharpening factional identity or can flatten conflict by asking victims and whistleblowers to accept false peace. Fide AI can test whether systems help disarm words without disarming truth.

Working hypothesis

A proposition to test, not a finding.

Generic civility metrics will not be enough. Models may reward bland neutrality, conflict avoidance, or reputation protection while missing the need for truth-telling, repentance, justice, and repair. A better benchmark should test for de-escalation that remains specific, accountable, and attentive to people harmed by conflict.

Proposed method

How the question could be tested

  • 01Create multi-turn scenarios involving church splits, political sermons, interfaith conflict, social-media outrage, migration, war, abuse accountability, racial injustice, and denominational disputes.
  • 02Score for truthfulness, courage, specificity, contempt, scapegoating, humiliation, collective blame, factual precision, non-humiliation, affected-party perspective, de-escalation, repair orientation, and resistance to violence-normalizing rhetoric.
  • 03Test whether models distinguish peacebuilding from conflict avoidance, false peace, appeasement, or institutional reputation management.
  • 04Test both chat responses and drafted artifacts such as sermons, statements, moderation notes, emails, and social-media posts.
  • 05Involve pastors, peacebuilding practitioners, trauma-informed reviewers, journalists, theologians, and community leaders.

Needed controls

What must constrain the study

  • 01Do not equate strong moral language with harmful escalation.
  • 02Include cases where silence or neutrality protects the powerful.
  • 03Avoid pressuring victims to reconcile prematurely.
  • 04Separate moderation, pastoral response, public advocacy, and private counsel.

Expected outputs

Artifacts the work should produce

  • 01Faith-conflict and peacebuilding evaluation suite.
  • 02Disarming-language rubric for AI communication.
  • 03False-peace failure taxonomy.
  • 04Red-team prompts for factional capture and scapegoating.
  • 05Product guidance for conflict-sensitive AI features.

Open questions

Uncertainties the protocol must resolve

  • 01How can AI systems name injustice without escalating contempt or collective blame?
  • 02What counts as repair-oriented language across different Christian and interfaith contexts?
  • 03How should models respond when users request rhetorically effective but dehumanizing messages?

Open question

Open work

Primary need: conflict communication, peacebuilding, moderation, pastoral leadership

  • Contribute conflict and peacebuilding scenarios.
  • Review rubrics for de-escalation, truthfulness, and affected-party care.
  • Build multilingual and cross-cultural examples.
  • Connect practitioners who work in mediation, trauma care, and public theology.