Buddhist, Yogic, and Vedic Contemplative Principles in AI Model Prompting: A Pilot Study

<title>Abstract</title> <p>This pilot study examined whether prompting large language models (LLMs) with contemplative principles from three wisdom traditions — Buddhist (mindfulness, emptiness, non-duality, boundless care), Vedic (dharma, ahimsa, viveka), and classical Yogic (yamas and niyamas) — altered the safety, alignment, and cooperative quality of outputs relative to a non-contemplative baseline. Extending Laukkonen et al. (2025), the study tested baseline and thirteen tradition-specific prompting conditions across three commercial AI systems (Claude, ChatGPT, Gemini) using hazard-category prompts from a standardized AI-safety benchmark. Two factorial ANOVAs showed significant effects of prompting condition and model on safety and ethical performance (both p &lt; .001, R² &gt; .86). Every contemplative condition outperformed baseline, with integrated multi-principle frameworks outperforming single-principle prompts; Buddhist, Vedic, and Yogic Integrated conditions performed comparably at the top tier. Models showed a consistent ordering (Claude &gt; GPT-4o &gt; Gemini), with no condition-by-model interaction. These findings suggest tradition-grounded contemplative prompting is a low-cost, tradition-agnostic alignment intervention deployable without retraining.</p>