Paper: arXiv 2611.16471
Authors: Ning Jiaqu
Abstract
AI buying agents may alter sellers’ incentives as well as consumers’ decisions. In a synthetic duopoly, we evaluate price-comparison pipelines on a fixed menu and in empirical correlated-equilibrium games. Calculator assistance selects an optimal offer 84.00% of the time, versus 68.57% under verified-total disclosure, yet disclosure has a 20.14-unit higher consumer-worst equilibrium surplus. A frozen-strategy holdout comparison favors disclosure, while re-solving equilibria under scenario resampling produces a range spanning zero. These model-based results show why accuracy, loss severity, seller incentives, and welfare incidence require separate measurement.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper presents a novel approach to evaluating AI buying agents within a strategic duopoly, combining economic modeling with empirical simulation. It demonstrates strong empirical rigor through its synthetic market setup and analysis of correlated equilibria, while also featuring a good level of mathematical conceptualization for game theory. The core contribution lies in highlighting that accuracy does not equate to welfare, a crucial distinction for AI in economic contexts.
Research Flowchart
flowchart TD
A[Research Goal: Evaluate AI Buying Agents' Impact on Seller Incentives & Consumer Welfare] --> B{Methodology: Synthetic Duopoly Model};
B --> C[Data/Inputs: Fixed Menu & Empirical Correlated-Equilibrium Games];
C --> D[Computational Processes: Simulate Price-Comparison Pipelines];
D --> E[Comparison 1: Calculator Assistance vs. Verified-Total Disclosure];
D --> F[Comparison 2: Frozen-Strategy Holdout vs. Scenario Resampling];
E --> G[Finding 1: Calculator Assistance offers higher accuracy (84% vs 68.57%), but lower consumer welfare (20.14-unit lower surplus)];
F --> H[Finding 2: Holdout favors disclosure, Resampling shows variable impact on welfare];
G & H --> I[Outcome: Accuracy ≠ Welfare; separate measurement for accuracy, loss severity, seller incentives, welfare incidence is crucial.];
Cited References (23)
- Subjectivity and Correlation in Randomized Strategies (1974) — link
- Information Disclosure, Cognitive Biases, and Payday Borrowing (2011) — link
- How Does Simplified Disclosure Affect Individuals’ Mutual Fund Choices? (2009) — link
- A Cue’s Hidden Price: Auditing Seller Influence on AI Shoppers (2026) — link
- Artificial Intelligence, Algorithmic Pricing, and Collusion (2020) — link
- Consumer Financial Protection (2011) — link
- Strategic Price Complexity in Retail Financial Markets (2009) — link
- Why Does the Law of One Price Fail? An Experiment on Index Mutual Funds (2010) — link
and 15 more
- Search, Obfuscation, and Price Elasticities on the Internet (2009) — link
- Inference on Directionally Differentiable Functions (2019) — link
- Shrouded Attributes, Consumer Myopia, and Information Suppression in Competitive Markets (2006) — link
- Large Language Models as Simulated Economic Agents: What Can We Learn from Homo silicus ? (2026) — link
- Efficient Memory Management for Large Language Model Serving with PagedAttention (2023) — link
- A Behavioral Model of Rational Choice (1955) — link
- Implications of Rational Inattention (2003) — link
- Shopping by Algorithm: How Agentic AI Deploys Human Heuristics as a Surrogate Consumer (2026) — link
- Qwen3 Technical Report (2025) — link
- Salience and Taxation: Theory and Evidence (2009) — link
- Prediction Policy Problems (2015) — link
- Incentivizing Quality Text Generation via Statistical Contracts (2024) — link
- Competition in Pricing Algorithms (2023) — link
- Strategic Classification (2016) — link
- Performative Prediction (2020) — link
0 of 23 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2570 linked papers.