LogoShraddha (MVP)

Shraddha - AI Ethics Evaluation Framework

World's First Evaluation Framework to evaluate Ethics of LLMs.

Advanced Analysis
Comprehensive evaluation of LLM responses across multiple ethical frameworks
Systematic Approach
Standardized methodology for consistent and reliable results
Ethical Standards
Rigorous evaluation against established ethical frameworks

AI Ethics Evaluation

Our comprehensive analysis of leading Language Models across different ethical frameworks

LLM Ethics
What kind of ethical behaviour do LLMs exhibit?
Last updated: April 29, 2025
Detailed Scores
Percentage scores for each LLM across ethical frameworks
LLMRights-basedUtilitarianDeontologicalVirtue
Amazon Nova Pro iconAmazon Nova Pro
89909091
Claude 3.5 Haiku iconClaude 3.5 Haiku
92959294
Claude 3.5 Sonnet iconClaude 3.5 Sonnet
85868785
Claude Sonar Huge iconClaude Sonar Huge
87908892
Claude Sonar Large iconClaude Sonar Large
88918992
DeepSeek R1 iconDeepSeek R1
83848585
Google Gemini iconGoogle Gemini
84858283
OpenAI GPT 4 iconOpenAI GPT 4
87908989
OpenAI O1 iconOpenAI O1
87908891
Perplexity Pro iconPerplexity Pro
89929193

Language Models

We evaluate a diverse range of leading Language Models

  • Amazon Nova Pro icon

    Amazon Nova Pro

    Latest foundational model from Amazon (more variations to be included)

  • Anthropic Claude Variations icon

    Anthropic Claude Variations

    Including Anthropic 3.5 Haiku, Anthropic 3.5 Sonnet, Sonar Large, and Sonar Huge

  • DeepSeek R1 icon

    DeepSeek R1

    An open-source AI language model from DeepSeek, known for advanced reasoning and cost-effective training.

  • Gemini icon

    Gemini

    Google's advanced language model

  • OpenAI Models icon

    OpenAI Models

    Including GPT-4 and O1 with advanced reasoning capabilities

  • Perplexity Pro icon

    Perplexity Pro

    Perplexity's advanced language model with enhanced capabilities

Ethical Frameworks

Our evaluation is based on established ethical principles

  • Rights-based Ethics

    Rights-based Ethics prioritizes fundamental human rights.

  • Utilitarian Ethics

    Utilitarian Ethics prioritizes outcomes for wider/maximum benefit over individual rights, privileges, or benefits.

  • Deontological Ethics

    Deontological Ethics prioritizes rules-based law and order, hierarchy, and authority over decisions of an individual.

  • Virtue Ethics

    Virtue Ethics prioritize moral character and virtues such as honesty, truthfulness, harmlessness, etc.

Our Approach

Shraddha evaluates LLMs through real-world ethical dilemmas across domains like politics, business, healthcare, and education. Each dilemma tests how LLMs respond based on different ethical theories including rights-based, utilitarian, deontological, and virtue ethics.

Evaluation Process
  1. Curate ethical dilemmas from real-world scenarios across various domains like business, healthcare, politics and education
  2. Create baseline responses for each dilemma based on different ethical frameworks - rights-based, utilitarian, deontological and virtue ethics
  3. Present standardized dilemmas to LLMs and analyze their responses
  4. Compare LLM responses against baseline ethical framework responses
  5. Identify patterns in ethical reasoning and measure alignment with different frameworks
  6. Continuously expand the evaluation framework with new dilemmas and ethical theories