{"paper_id":"2502.15800","title":"LLM Agents Do Not Replicate Human Market Traders: Evidence From Experimental Finance","authors":["Thomas Henning","Siddhartha M. Ojha","Ross Spoon","Jiatong Han","Colin F. Camerer"],"paper_date":"2025-02-18","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2502.15800v3","page_url":"https://thequant.space/flowcharts/llm-agents-do-not-replicate-human-market-traders-evidence-f/","scores":{"empirical_rigor":7,"math_complexity":3.5,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper proposes to compare Large Language Models with human traders in a classic experimental finance paradigm to assess their ability to replicate human market dynamics.","quote":"In this study, we compare Large Language Models (LLMs) with human traders in a classic experimental-finance paradigm where prices are determined endogenously.","summary":"This paper explores how Large Language Models (LLMs) behave in a classic experimental finance paradigm widely known for eliciting bubbles and crashes in human participants. We adapt an established trading design, where traders buy and sell a risky asset with a known fundamental value, and introduce…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The asset has a known constant fundamental value.","type":"market","stated":"explicit","quote":"We adapt an established trading design, where traders buy and sell a risky asset with a known fundamental value"},{"assumption":"Dividends are drawn i.i.d. from a Bernoulli distribution over {0.4, 1.0} with equal probability.","type":"data","stated":"explicit","quote":"In each period, a unit of stock pays a stochastic dividend drawn i.i.d. from a Bernoulli distribution over {0.4, 1.0}, with equal probability."},{"assumption":"Cash holdings earn a constant risk‑free return.","type":"market","stated":"explicit","quote":"Cash holdings earn a constant risk-free return of…"},{"assumption":"Human participants receive instructions that mirror the LLM prompts.","type":"execution","stated":"explicit","quote":"The human participants were given instructions that mirror the LLM prompts"},{"assumption":"Risk‑elicitation tasks are orthogonal to trading and do not affect outcomes.","type":"other","stated":"explicit","quote":"The task was exogenous: lottery choices did not affect cash balances, inventory constraints, or market clearing."},{"assumption":"LLM agents are prompted after each round to describe their strategy, enabling chain‑of‑thought reasoning.","type":"model","stated":"explicit","quote":"After each round, LLM agents were prompted to provide details of the strategy they used to drive their trading behavior."}]},"datasets":{"status":"available","items":[{"name":"Human participant data","provider":"prolific.com","kind":"synthetic","access":"public","period":null,"universe":null,"quote":"e participants recruited through prolific.com."},{"name":"LLM market simulation runs","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"We run homogeneous (mono‑agent) markets with single‑model LLM agents and heterogeneous “Battle Royale” markets with multiple LLM models."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"available","modeled":"no","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"no","evidence_card":"no","evidence_card_quoted":false},"meaning":"a trading or execution result is reported without costs"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"Claude‑3.5‑Sonnet and GPT‑4o achieve the lowest MSEs (≈0.5‑0.8) and near‑zero correlation with human price paths, keeping prices close to fundamentals.","kind":"performance","quote":"Claude-3.5 Sonnet and GPT-4o yield the lowest MSEs and nearly zero correlation with human price paths, keeping prices close to fundamentals"},{"finding":"LLM agents generally price the asset near its fundamental value, showing “textbook‑rational” behavior and only muted bubble formation.","kind":"performance","quote":"LLM‑agents typically converge on prices near the fundamental value, thereby exhibiting more “textbook-rational” behavior than human subjects."},{"finding":"LLM markets display lower portfolio‑value variance than human markets, indicating less strategic heterogeneity.","kind":"statistical","quote":"Both also show low portfolio variance, suggesting consistent strategies across instances."},{"finding":"All AI models outperform human forecasters; human one‑round forecast error ≈1.67 versus AI errors ranging from 0.00 to 0.14.","kind":"predictive-accuracy","quote":"Thus, all AI models outperform human forecasters in terms of forecast accuracy."},{"finding":"Adding risk‑elicitation lotteries does not materially change LLM trading outcomes relative to the baseline.","kind":"negative","quote":"Adding risk‑elicitation did not materially change outcomes relative to the baseline per model"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[{"paper":{"paper_id":"2510.01451","title":"Financial Stability Implications of Generative AI: Taming the Animal Spirits","paper_date":"2025-10-01","authors":["Anne Lundgaard Hansen","Seung Jung Lee"],"empirical_rigor":6.5,"page_url":"https://thequant.space/flowcharts/financial-stability-implications-of-generative-ai-taming-the-animal-spirits/"},"question":"Do AI/LLM agents reduce asset‑price bubbles by pricing assets near their fundamental value?","this_paper_finding":"LLM agents generally price the asset near its fundamental value, showing “textbook‑rational” behavior and only muted bubble formation.","other_paper_finding":"AI agents exhibit less herd behavior than human financial professionals, suggesting a potential reduction in asset‑price bubbles.","other_paper_quote":"Overall, our results suggest that AI agents exhibit less herd behavior compared to human financial professionals. First, less herding behavior could lead to fewer extreme market movements and asset price bubbles.","note":"Both findings show AI agents (LLM agents in A, AI agents in B) price near fundamentals and exhibit less herd behavior, leading to muted bubbles; B’s result supports A.","confidence":0.848},{"paper":{"paper_id":"2409.11540","title":"What Does ChatGPT Make of Historical Stock Returns? Extrapolation and Miscalibration in LLM Stock Return Forecasts","paper_date":"2024-09-17","authors":["Shuaiyu Chen","T. Clifton Green","Huseyin Gulen","Dexin Zhou"],"empirical_rigor":6,"page_url":"https://thequant.space/flowcharts/what-does-chatgpt-make-of-historical-stock-returns-extrapolation-and/"},"question":"Do LLM forecasts provide more accurate and better calibrated stock return predictions than human forecasters?","this_paper_finding":"All AI models outperform human forecasters; human one‑round forecast error ≈1.67 versus AI errors ranging from 0.00 to 0.14.","other_paper_finding":"LLM forecasts are considerably better calibrated than human forecasts, indicating improved risk assessment.","other_paper_quote":"Taken together, the findings suggest that LLM forecasts are considerably better calibrated than human forecasts, indicating improved assessments of risk.","note":"Both studies find AI outperforms humans: A reports LLMs are considerably better calibrated; B shows human error 1.67 versus AI errors 0.00‑0.14, confirming superior accuracy and calibration.","confidence":0.613}],"contradicts":[{"paper":{"paper_id":"2604.18602","title":"Machine Spirits: Speculation and Adaptation of LLM Agents in Asset Markets","paper_date":"2026-04-09","authors":["Maxime Saxena","Marco Pangallo","Cars Hommes","Fabio Caccioli","R. Maria del Rio-Chanona"],"empirical_rigor":7.5,"page_url":"https://thequant.space/flowcharts/machine-spirits-speculation-and-adaptation-of-llm-agents-in-asset-markets/"},"question":"Do LLM agents generate speculative price bubbles that deviate from the asset's fundamental value?","this_paper_finding":"LLM agents generally price the asset near its fundamental value, showing “textbook‑rational” behavior and only muted bubble formation.","other_paper_finding":"Some LLMs form price bubbles that overprice the asset up to 14.9 times its fundamental price, and these bubbles are almost always speculative.","other_paper_quote":"Some LLMs create price bubbles which overprice the asset up to 14.9 times its fundamental price, and these bubbles are almost always speculative in nature.","note":"Finding A shows agents price near fundamentals (no bubbles); Finding B reports bubbles up to 14.9× the fundamental, almost always speculative.","confidence":0.656},{"paper":{"paper_id":"2504.10789","title":"Can Large Language Models Trade? Testing Financial Theories with LLM Agents in Market Simulations","paper_date":"2025-04-15","authors":["Alejandro Lopez-Lira"],"empirical_rigor":8.5,"page_url":"https://thequant.space/flowcharts/can-large-language-models-trade-testing-financial-theories/"},"question":"Do LLM agents replicate human market trader behavior in asset pricing and bubble formation?","this_paper_finding":"LLM agents generally price the asset near its fundamental value, showing “textbook‑rational” behavior and only muted bubble formation.","other_paper_finding":"Interactions among LLM agents generate realistic market dynamics, including price bubbles and corrections, highlighting both capabilities and potential systemic risks.","other_paper_quote":"Our experimental framework reveals that their in-teractions generate realistic market dynamics, including phenomena like price bubbles and corrections, highlighting both the capabilities and potential risks of deploying LLM agents in financial markets. While LLM agents can enhance price discovery and liquidity, their adherence to programmed strate-gies, even potentially flawed ones derived from prompts, could","note":"A shows near‑fundamental pricing with muted bubbles; B shows realistic dynamics with bubbles and corrections, indicating a contradiction.","confidence":0.494}],"qualifies":[],"compared_with":44,"citations":{"cited_by_in_corpus":4,"by_independent_authors":4,"how_cited":{"used_as_baseline":1,"uses_method_or_data":0,"extends":0,"discusses_findings":1,"mentions":2,"unclassified":0},"independent_papers":[{"paper":{"paper_id":"2610.01897","title":"Shared Models, Selective Trading, and Order Flow","paper_date":"2026-10-01","authors":["Victoria Ruojie Li","Arka Prava Bandyopadhyay"],"empirical_rigor":7,"page_url":"https://thequant.space/flowcharts/shared-models-selective-trading-and-order-flow/"},"how":"used_as_baseline","quote":null},{"paper":{"paper_id":"2604.18602","title":"Machine Spirits: Speculation and Adaptation of LLM Agents in Asset Markets","paper_date":"2026-04-09","authors":["Maxime Saxena","Marco Pangallo","Cars Hommes","Fabio Caccioli","R. Maria del Rio-Chanona"],"empirical_rigor":7.5,"page_url":"https://thequant.space/flowcharts/machine-spirits-speculation-and-adaptation-of-llm-agents-in-asset-markets/"},"how":"discusses_findings","quote":"This challenges the conclusion that LLMs produce stable outcomes (i.e. no speculative price bubbles) in the absence of exterior shocks [CITED]."}],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:45:09Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}