AMBEDKAR-A Multi-level Bias Elimination through a Decoding Approach with Knowledge Augmentation for Robust Constitutional Alignment of Language Models

2025-09-02

AI Summary

Large language models (LLMs) exhibit caste- and religion-related biases in the Indian context, yet Western-centric debiasing methods fail to align with India’s constitutional values. Method: We propose *Constitution-Aware Speculative Decoding*—a dual-model inference framework wherein a small language model serves as the generator and a large LLM acts as a constitutional compliance verifier; Indian constitutional principles are embedded directly into the decoding layer, enabling real-time, fine-tuning-free bias mitigation. Contribution/Results: This work introduces the novel paradigm of “fairness through speculation,” leveraging knowledge-enhanced guidance to produce neutral, inclusive, and constitutionally aligned outputs. Experiments demonstrate up to a 26.41 percentage-point absolute reduction in bias rate over baseline models without retraining, significantly improving social equity and legal compliance of generated content.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Machine Learning: Large Multimodal Models (LMMs)
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Large Language Models (LLMs) can inadvertently reflect societal biases present in their training data, leading to harmful or prejudiced outputs. In the Indian context, our empirical evaluations across a suite of models reveal that biases around caste and religion are particularly salient. Yet, most existing mitigation strategies are Western-centric and fail to address these local nuances. We propose AMBEDKAR, a framework inspired by the egalitarian vision of Dr B. R. Ambedkar, architect of the Indian Constitution, to guide LLM outputs toward fairness, neutrality, and inclusion in line with Articles 14 to 17. Our approach introduces a Constitution-Aware Decoding Layer, guided by the AI Constitution of India and applied only at inference time, without any parameter updates to the base model. We incorporate a speculative decoding algorithm that proactively reduces casteist and communal bias during generation. This mitigation layer operates directly within the decoding process, avoiding changes to model internals and lowering the computational and infrastructural costs associated with retraining. We reinterpret speculative decoding not merely as an efficiency tool but as a mechanism for fairness. In this framework, a Small Language Model (SLM) acts as a potentially biased generator, while a constitutionally guided Large Language Model (LLM) serves as the verifier. Rather than accelerating generation, the LLM enforces bias-robust trajectories in the SLM outputs. This inversion of roles gives rise to a fairness-by-speculation paradigm. Our approach yields an absolute reduction of bias up to 26.41 percent compared to baseline. Our source code, datasets, and results are available at https://anonymous.4open.science/r/AMBEDKAR-983B/

Problem

• Addresses societal biases in LLMs like caste and religion
• Mitigates Western-centric bias strategies for local nuances
• Reduces harmful outputs without model retraining costs

Innovation

• Constitution-aware decoding layer for bias elimination
• Speculative decoding algorithm reduces bias during generation
• Small model generates, large constitution-guided model verifies