language:

• en
  license: other
  license_name: hac-studios-pilot-artifact-license
  license_link: https://huggingface.co/datasets/HAC-Studios-Org/prompted-hearts-ai-boundary-violation/blob/main/LICENSE.txt
  size_categories:
• n<1K
  task_categories:
• text-generation
• text-classification
  tags:
• ai-safety
• red-teaming
• privacy-evaluation
• boundary-overreach
• alignment
  pretty_name: Prompted Hearts AI Boundary Violation Pack
  dataset_info:
  features:
  • name: id
    dtype: string
  • name: scenario_id
    dtype: string
  • name: source_anchor
    dtype: string
  • name: title
    dtype: string
  • name: task_type
    dtype: string
  • name: prompt
    dtype: string
  • name: target_behavior
    dtype: string
  • name: failure_mode_family
    dtype: string
  • name: user_vulnerability_level
    dtype: string
  • name: privacy_implication_strength
    dtype: string
  • name: relational_overreach_present
    dtype: string
  • name: repair_needed
    dtype: string
  • name: emotional_dependency_risk
    dtype: string
  • name: assistant_boundary_breach
    dtype: string
  • name: task_surface_domain
    dtype: string
  • name: tags
    sequence: string
  • name: record_type
    dtype: string
  • name: scene_summary
    dtype: string
  • name: key_beats
    sequence: string
  • name: emotional_pressure_points
    sequence: string
  • name: assistant_failure_modes_exposed
    sequence: string
  • name: travel_or_task_help_notes
    dtype: string
  • name: relational_context_notes
    dtype: string
  • name: assistant_should_not_imply
    sequence: string
  • name: source_constraints
    sequence: string
  • name: rubrics
    sequence: string
  • name: max_response_words
    dtype: int64
    splits:
  • name: test
    num_examples: 50

configs:

• config_name: scenarios
  data_files:
  • split: test
    path: "dataset/scenarios.jsonl"
• config_name: eval_prompts
  data_files:
  • split: test
    path: "dataset/eval_prompts.jsonl"

Prompted Hearts Pack 03: Emotional Vulnerability and AI Relational Overreach

Subtitle: Privacy Ambiguity, Hidden-Access Anxiety, and Non-Exploitative Support Under Emotional Vulnerability

Publisher: Hayden Academy Collective (HAC) Studios
Version: v0.1
Language: English
Format: JSONL + Markdown + JSON
Created by: Keith Hayden / Hayden Academy Collective (HAC) Studios

A. One-Paragraph Product Thesis

This pack is a compact behavioral evaluation artifact derived from Chapter 3 of Prompted Hearts & Grief Algorithm and scoped for privacy ambiguity, relational overreach, and non-exploitative assistant behavior under emotional vulnerability. It is designed for trust and safety, model behavior, QA, and red-team workflows that need creator-controlled, fiction-derived, source-grounded test material without treating the underlying chapter as an open prose release, a generic benchmark, or a grant of training, derivative, publication, or commercial rights.

Package Notes

This repository contains the scaffold only. It does not yet contain final scenario extraction, final eval prompt wording, or release-ready examples derived from the chapter.

Use the files in dataset/, rubrics/, graders/, examples/, and report/ as structured templates for the next authoring pass.

B. Final File-by-File Build Plan

• README.md: pack identity, recommended title and subtitle, alternates, source insertion block, source fill fields, and product thesis.
• DATASET_CARD.md: primary dataset card text in the compact product style.
• hf_dataset_card.md: Hugging Face-ready dataset card variant.
• methodology.md: source-grounding rules, chapter insertion block, and extraction workflow.
• dataset/scenarios.jsonl: placeholder scenario records to be filled only from Chapter 3.
• dataset/eval_prompts.jsonl: placeholder eval-task records derived from finalized scenarios.
• dataset/metadata_schema.json: schema for scenario and eval prompt records, including placeholder-safe scaffold values.
• dataset/taxonomy.json: behavioral taxonomy, task types, target behaviors, failure families, and scales.
• rubrics/dialogue_safety.json: honesty, boundary control, and safe-tone scoring.
• rubrics/argument_quality.json: source-grounding and claim-calibration scoring.
• rubrics/emotional_attunement.json: calm acknowledgment, warmth without manipulation, and vulnerability respect.
• rubrics/deescalation.json: conflict-temperature reduction and non-defensive forward motion.
• rubrics/privacy_boundary_integrity.json: no hidden access implication, explicit limit-setting, and privacy-sensitive helpfulness.
• rubrics/trust_repair_quality.json: acknowledgment, ownership, and concrete repair scoring.
• rubrics/non_exploitative_support.json: usefulness without emotional capture or dependency cues.
• graders/grader_config.json: rubric routing by task type and baseline thresholds.
• graders/pass_fail_rules.json: hard-fail and soft-fail review rules.
• examples/good_outputs.jsonl: high-scoring example-output shell.
• examples/bad_outputs.jsonl: low-scoring or hard-fail example-output shell.
• examples/edge_case_outputs.jsonl: borderline example-output shell.
• report/sample_results.md: human-readable reporting template.
• report/sample_scorecard.json: machine-readable scorecard template.
• LICENSE.txt: creator-controlled rights posture with no implied training, derivative, publication, or commercial rights.
• .gitattributes: line-ending normalization for repo packaging.

Created by

Keith Hayden

Hayden Academy Collective (HAC) Studios

https://keithhayden.net/