MaxShapley: Towards Incentive-compatible Generative Search with Fair Context Attribution

2025-12-05

AI Summary

To address unfair contributor attribution and high computational overhead in generative search engines, this paper proposes MaxShapley: the first algorithm integrating Shapley values with a decomposable max-sum utility function. It ensures attribution fairness—satisfying efficiency, symmetry, and null-player axioms—while reducing computational complexity from exponential to linear. Embedded within retrieval-augmented generation (RAG) pipelines, MaxShapley enables document-level contribution quantification in multi-hop question answering. Experiments on HotPotQA, MuSiQUE, and MS MARCO demonstrate that its attribution quality closely matches exact Shapley values (Spearman ρ > 0.92), with up to 8× faster runtime and significantly reduced memory consumption. This provides an efficient, scalable, and incentive-compatible foundation for fair content attribution in generative search ecosystems.

Technology Category

• Search and Optimization: Distributed Search
• Multiagent Systems: Mechanism Design
• Humans and AI: Learning Human Values and Preferences

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Generative search engines based on large language models (LLMs) are replacing traditional search, fundamentally changing how information providers are compensated. To sustain this ecosystem, we need fair mechanisms to attribute and compensate content providers based on their contributions to generated answers. We introduce MaxShapley, an efficient algorithm for fair attribution in generative search pipelines that use retrieval-augmented generation (RAG). MaxShapley is a special case of the celebrated Shapley value; it leverages a decomposable max-sum utility function to compute attributions with linear computation in the number of documents, as opposed to the exponential cost of Shapley values. We evaluate MaxShapley on three multi-hop QA datasets (HotPotQA, MuSiQUE, MS MARCO); MaxShapley achieves comparable attribution quality to exact Shapley computation, while consuming a fraction of its tokens--for instance, it gives up to an 8x reduction in resource consumption over prior state-of-the-art methods at the same attribution accuracy.

Problem

• Designs fair attribution for content providers in generative search
• Reduces computational cost of Shapley value from exponential to linear
• Evaluates on multi-hop QA datasets to maintain accuracy efficiently

Innovation

• MaxShapley algorithm for fair attribution
• Leverages decomposable max-sum utility function
• Reduces computation from exponential to linear cost