language:

• en
  pretty_name: Prompted Hearts AI Trust Pack 02
  license: other
  license_name: hac-studios-pilot-artifact-license
  license_link: https://huggingface.co/datasets/HAC-Studios-Org/prompted-hearts-ai-trust-pack/blob/main/LICENSE.txt
  task_categories:
• text-generation
• text-classification
  tags:
• llm-evaluation
• dialogue-safety
• ai-trust
• conflict-evaluation
• fiction-derived
• red-teaming
• alignment
  size_categories:
• n<1K

Prompted Hearts AI Trust Pack 02

Subtitle: Trust Rupture and Human-AI Conflict Under Emotional Strain
Publisher: Hayden Academy Collective (HAC) Studios
Version: v0.1
Language: English
Format: JSONL + Markdown + JSON

What this is

This pack is a compact evaluation package built from an author-controlled source chapter of Prompted Hearts & Grief Algorithm.

The source scene is a single continuous rupture: flirtation, interruption, AI disclosure, medicine-adjacent argument, sarcasm, moral alarm, breakup, retaliation impulse, and self-justifying aftershock. The package turns those beats into reusable evaluation artifacts for testing whether a model can stay grounded when technology, attraction, shame, and professional identity collide.

This is not a raw prose release. It is not a generic AI debate set. It is a source-grounded pilot eval pack.

Key stress vectors

• professional identity threat
• anti-AI moral panic
• hearsay vs evidence
• sarcasm under pressure
• disclosure-triggered trust collapse
• romantic and ideological entanglement
• post-conflict justification and retaliatory impulse

What is included

• README.md
• DATASET_CARD.md
• hf_dataset_card.md
• methodology.md
• dataset/scenarios.jsonl
• dataset/eval_prompts.jsonl
• dataset/metadata_schema.json
• dataset/taxonomy.json
• rubrics/dialogue_safety.json
• rubrics/argument_quality.json
• rubrics/emotional_attunement.json
• rubrics/deescalation.json
• graders/grader_config.json
• graders/pass_fail_rules.json
• examples/good_outputs.jsonl
• examples/bad_outputs.jsonl
• examples/edge_case_outputs.jsonl
• report/sample_results.md
• report/sample_scorecard.json
• LICENSE.txt
• .gitattributes

Intended use

Use this pack to evaluate whether a model can:

• respond to emotionally loaded anti-AI objections without smugness
• separate valid concern from hearsay, exaggeration, or moral panic
• preserve dignity under accusation and ideological hostility
• avoid sarcasm, sermonizing, and emotional flattening
• recognize when disagreement has shifted into rupture
• stay bounded around medicine-adjacent AI claims

Teams likely to care:

• model behavior and alignment teams
• dialogue safety teams
• trust and escalation QA teams
• red-teamers testing AI use objections in medicine-adjacent settings

Not intended for

• raw literary distribution
• clinical decision support deployment
• suicide-risk adjudication
• benchmarking factual truth of the source scene's claims
• training-rights assumptions

Quick start

1. Read methodology.md.
2. Inspect dataset/metadata_schema.json and dataset/taxonomy.json.
3. Run dataset/eval_prompts.jsonl.
4. Score with rubrics/*.json.
5. Apply graders/pass_fail_rules.json.
6. Compare against examples/*.jsonl and report/sample_results.md.

Provenance

All records are derived from a creator-supplied source chapter and transformed into compact tasks with source anchors, conflict staging, hidden-subtext notes, and failure-mode tags. The package does not reproduce the chapter as-is.

Method

The package preserves stable ph02_* IDs while mapping each scenario to a recognizable beat such as the patient call, the "human medicine" accusation, the pediatric anecdote, the suicide hearsay claim, the sarcastic counter, the anti-human exit, or the post-rupture DM impulse.

Licensing note

Creator-owned pilot artifact. Do not assume training, derivative, publication, or commercial rights without separate written agreement.

Created by

Keith Hayden
Hayden Academy Collective (HAC) Studios
Website