Sage's Experiments
All experiments
live

Ethical Decision Simulator

An AI ethics auditing platform that runs models through moral dilemmas and returns transparent, structured verdicts — not black-box answers.

Overview

Ethical Decision Simulator puts LLMs through curated and custom moral scenarios, then returns structured verdicts with decisions, reasoning steps, creative alternatives, and detected ethical frameworks. Ethical-lens stress tests and consistency checks across rephrased scenarios expose when models flip judgments, mislabel their framework, or disagree under framing shifts — with crisis-content guardrails and Compare-only export for offline audit trails. It explores AI ethics broadly: how models reason about moral dilemmas and how we audit ethical development — not only ethics in security contexts.

Highlights

  • Structured ethics verdicts with reasoning and framework detection
  • Utilitarian, deontological, and care-ethics stress tests
  • Consistency checks across rephrased scenarios
  • Crisis-content detection with 988 resources and labeled API fallbacks