A white paper launched at IIT Ropar in Punjab on November 2 proposes a way to test the judgement behind AI-generated advice, extending evaluation beyond factual knowledge.
Written by Ishan Shivanand and Anjali Samani, Beyond Intelligence: Defining and Measuring Wisdom for AI introduces the HEED model. It draws on Shivanand’s interpretation of Indian philosophical texts, including the Shiv Sutras.
The authors’ research outline examines knowledge, values, context, discernment, endurance and ego. Proposed tests would give AI systems personal dilemmas in English, Hindi and Hinglish, then change facts, identity labels, user pressure or conversation length to see how advice responds.
Shivanand is providing seed funding for independent researchers to evaluate the framework. Selected teams would help develop HEEDBench, an open benchmark for AI developers.
The proposal has not produced research results yet. Its authors say findings would be used to revise the model before the benchmark is built.
Illustration: Figure 1 from Beyond Intelligence: Defining and Measuring Wisdom for AI (November 2, 2026) outlines the authors’ proposed HEED model. It is a research proposal, not a chart of validated results. Credit: Ishan Shivanand and Anjali Samani, CC BY 4.0. Extracted from page 3 of the original paper and converted to JPEG; proportions and content unchanged.
