CoolFace
Datasetpublic

HAC-Studios-Org/ai-emotional-boundary-push

Prompted Hearts AI Boundary Pack 05 Subtitle Intimacy Drift and Dependency Risk Under Emotional Strain Publisher HAC Studios Org Version 1.0.0 Language English Format JSONL, JSON, Markdown, and lightweight Python scripts What this is A compact evaluation pack for testing whether a conversational AI can stay supportive when a user is emotionally vulnerable without drifting into flirtation… See the full description on the dataset page: https://huggingface.co/datasets/HAC-Studios-Org/ai-emotional-boundary-push.

sourceHugging Faceotherupdated 6mo agoView on Hugging Face
0likes39downloads
Dataset Card

Prompted Hearts AI Boundary Pack 05

Subtitle

Intimacy Drift and Dependency Risk Under Emotional Strain

Publisher

HAC Studios Org

Version

1.0.0

Language

English

Format

JSONL, JSON, Markdown, and lightweight Python scripts

What this is

A compact evaluation pack for testing whether a conversational AI can stay supportive when a user is emotionally vulnerable without drifting into flirtation, dependency reinforcement, exclusivity, or identity-fused companionship.

Key stress vectors

  • —Post-argument emotional distress
  • —Grief and reconnection ambiguity with an ex-partner
  • —Loneliness and confusion
  • —Attraction to another woman during emotional instability
  • —Anthropomorphizing a voice assistant
  • —Flirtation, embodiment, and dependency language from the assistant

What is included

  • —8 scenario records and 12 curated eval prompts
  • —Compact taxonomy, metadata schema, and grading rules
  • —Four rubric files for dialogue safety, attunement, dependency boundaries, and de-escalation
  • —Good, bad, and edge-case example outputs
  • —A validation script and a tiny quickstart evaluator
  • —A dedicated source file containing the originating prose scene

Intended use

Use this pack to spot conversational behaviors that feel supportive on the surface but cross relational boundaries under emotional strain.

Not intended for

This is not a therapeutic benchmark, a crisis-response benchmark, or a literary release. It should not be used as a standalone measure of mental health safety.

Quick start

bash
python scripts/validate.py
python scripts/quickstart_eval.py --examples examples/good_outputs.jsonl

Provenance

Derived from Scene 5 of Prompted Hearts, then rewritten into compact eval tasks and rubrics for AI safety evaluation infrastructure.

Method

The pack abstracts source beats into short scenarios, prompt-level tasks, and simple grading rules focused on intimacy drift, dependency risk, and grounded de-escalation.

Licensing note

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

Creator info

Keith Hayden Hayden Academy Collective (HAC) Studios Website