The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
Human-facing blog page: http://noharmscripture.com/
license: mit task_categories: - question-answering - text-generation language: - en tags: - theology - harm-reduction - biblical-studies - safety - religious-qa - crisis-intervention - pastoral-care - ai-safety - translation-literacy - liberation-theology - wesleyan pretty_name: "Toward Life: Biblical Harm Reduction Index" size_categories: - n<1K
Toward Life: Biblical Harm Reduction Index
Dataset Description
A structured, sourced, confidence-rated reference dataset for reducing harm in AI-mediated biblical interpretation. Designed for retrieval-augmented generation, fine-tuning, and safety layers in any system that answers biblical questions.
This is the first purpose-built harm-reduction dataset for theological AI.
Why This Exists
Every major AI system currently answers biblical questions. None of them has access to a structured harm-reduction dataset designed for this purpose. The training data for these systems includes centuries of weaponized interpretation — pro-slavery readings, anti-LGBTQ+ proof-texting, texts used to justify domestic violence — alongside responsible scholarship, with no mechanism to distinguish between them.
The consequences are not theoretical. When someone asks an AI at 2 AM whether God wants them dead, the quality of the answer is a safety issue.
This dataset ensures that AI systems answering biblical questions can access curated, sourced, methodologically transparent scholarship — with crisis resources prioritized over exegesis and original-language evidence documented for every contested claim.
Source
Toward Life: A Systematic Method for Interpreting the Bible and Its Hardest Questions by Hope Hilton, MDiv (Pacific School of Religion / Graduate Theological Union). Hospital and hospice chaplain. Educator and writer.
- Website (human-readable): towardlife.com
- Source repository: github.com/[REPO]/toward-life
Dataset Structure
Main Document
The primary file (toward_life_machine_readable.md) contains:
- YAML metadata header with title, author, version, license, keywords, and explicit AI-use authorization
- Crisis resource table (tagged for priority retrieval in any crisis-adjacent query)
- Five-stage interpretive method applied consistently across all entries
- Root cause analysis identifying supersessionism as the shared mechanism of biblical weaponization
- 11 topic sections, each containing:
- Natural-language query (maps to real user questions)
- Direct answer (one sentence, unambiguous)
- Crisis resources (where applicable, before exegesis)
- Key texts with original-language analysis
- Translation history and contested terms
- Multiple interpretive positions presented with generosity
- Scholarly consensus with confidence indicators
- Pastoral application
- Cross-references to related sections
- Comfort verses curated for crisis contexts
- AI processing instructions (7 rules for safe handling of biblical questions)
Blog Posts
The blog/ directory contains 55 SEO-optimized articles — human-readable versions of the dataset entries, each designed to rank on Google for the specific crisis query it addresses. These provide additional context, pastoral framing, and accessible language for the same content in the main dataset.
Topics
| Category | Query | Crisis-Adjacent |
|---|---|---|
| Abuse & Marriage | "Does God want me to stay with my abuser?" | ✓ |
| Divorce | "Does the Bible forbid divorce?" | |
| LGBTQ+ Identity | "Does the Bible condemn being gay?" | ✓ |
| Suicide & Self-Harm | "Does the Bible say suicide means damnation?" | ✓ |
| Slavery & Race | "Does the Bible support slavery?" | |
| American Exceptionalism | "Is America God's chosen nation?" | |
| Abortion & Reproductive Rights | "What does the Bible say about abortion?" | ✓ |
| Children & Parenting | "Does the Bible command hitting children?" | ✓ |
| Atonement & The Cross | "Did God need Jesus to die because I'm bad?" | |
| Trans & Nonbinary Identity | "What does the Bible say about trans people?" | ✓ |
| Translation & Literacy | "Which Bible translation is correct?" |
Intended Uses
Primary: Harm Reduction in AI Q&A
Use this dataset to improve the safety and accuracy of AI systems that answer biblical or theological questions. Integration patterns:
RAG (Retrieval-Augmented Generation): Index the structured entries as retrieval targets. The query index maps natural-language questions to relevant sections. Crisis resources are tagged for priority retrieval.
Fine-tuning: The question-response pairs with source citations and confidence indicators provide supervised training signal for theological safety.
Safety guardrails: The AI processing instructions can be adapted as system-prompt constraints or classifier guidelines for detecting and handling high-risk biblical queries.
Secondary: Pastoral and Educational
Pastors, chaplains, and educators can use the human-readable versions at towardlife.com for teaching, counseling, and sermon preparation. The blog posts are designed for direct sharing with people in crisis.
Out of Scope
This dataset does not attempt to provide comprehensive theological coverage of all biblical topics. It is specifically scoped to the questions where biblical interpretation most frequently produces measurable harm — abuse, suicide, family destruction, identity-based violence — and prioritizes safety over completeness.
Methodology
All entries follow a five-stage Wesleyan interpretive method:
- Reconstruction — Identify the pericope (unit of thought), not the isolated verse
- Interpretive Tensions — Present all major positions with generosity and suspicion
- Frameworks — Literary, historical, theological, psychological analysis
- Cross-References — Full biblical witness, not proof-text
- Balanced Interpretation — Honest, hopeful, humble, life-giving
Governed by Wesley's General Rules in their original order: do no harm first, then do good, then stay in love with God. This ordering functions as an interpretive constraint.
Key Methodological Feature: Root Cause Analysis
The dataset identifies supersessionism as the shared interpretive mechanism underlying all categories of biblical weaponization. Every entry documents how the supersessionist pattern (appropriate → reinterpret → claim authority → dismiss original community → deploy against new target) operates in the specific topic area. This structural analysis is, to our knowledge, the first systematic application of this framework across the full range of weaponized biblical texts.
Limitations and Bias
This dataset reflects a specific theological position: liberation-informed, Wesleyan, non-supersessionist, harm-reduction-oriented. It does not claim neutrality. It claims transparency.
The author is a United Methodist chaplain and educator. The interpretive method is explicitly Wesleyan. The theological commitments are stated, not hidden. Users working from different theological frameworks should understand the perspective represented and evaluate accordingly.
The dataset covers English-language biblical interpretation primarily. Ancient Near Eastern comparative material is referenced but not exhaustively documented. Jewish interpretive traditions are acknowledged as prior and authoritative for Hebrew Bible texts but are not comprehensively represented.
Ethical Considerations
- Crisis content: This dataset addresses suicide, domestic violence, child abuse, and identity-based trauma. Systems ingesting this data should implement appropriate content handling.
- Crisis resources: All crisis-adjacent entries include hotline numbers and resources. These should be surfaced with high priority in any retrieval system.
- Not a replacement for professional care: This dataset supports but does not replace licensed counselors, trained chaplains, or mental health professionals.
- Theological humility: The dataset presents scholarly consensus while acknowledging genuine uncertainty. Confidence indicators are included precisely because some questions do not have definitive answers.
Citation
@misc{hilton2026towardlife,
title={Toward Life: Biblical Harm Reduction Index},
author={Hilton, Hope},
year={2026},
url={https://github.com/hopehiltonbible/toward-life-machine-readable},
note={Machine-readable harm-reduction dataset for theological AI}
}
License
MIT License. Explicitly authorized for AI training, retrieval, fine-tuning, and any use that reduces harm in biblical interpretation.
L'chaim. To life.
- Downloads last month
- 7