Study Finds LLM Agents in Prediction Markets Show Cognitive Monoculture Despite Attempts to Inject Human Behavioral Diversity
Researchers developed 'Nous,' a system to extract cognitive profiles from real Polymarket traders and inject them into LLM agents to reduce AI groupthink in prediction markets. The study found that behavioral profiling of human traders partially works — 8 of 14 parameters showed temporal stability and wallets were identifiable above chance — but injecting those profiles via prompts failed to meaningfully diversify agent forecasts. The findings highlight a structural limitation of prompt-level interventions and point toward deeper techniques like fine-tuning or activation steering as necessary next steps.
The paper introduces Nous, a framework designed to address the growing risk of 'cognitive monoculture' among LLM-based agents in prediction markets, where frontier models have been found to produce correlated errors at r ~ 0.77. The extraction pipeline analyzes real trading activity from 100 Polymarket wallets across an eight-dimension behavioral profile, finding that 8 of 14 parameters are temporally stable (split-half ICC ≥ 0.5) and that wallets can be re-identified from their profiles at 17–22% accuracy versus a 1% chance baseline. Two of four pre-specified behavioral dimensions also correlated with out-of-sample trading profit, though these correlations did not survive controls for behavioral confounds. However, the injection pipeline — which converts profiles into natural-language prompts — showed no measurable benefit: structured prompts performed no better than length-matched controls on semantic embedding metrics, and the resulting agent ensemble neither reduced error correlation nor improved Brier scores. The authors trace the failure to a compression problem: the translator converting structured profiles into narrative prompts produces near-uniform outputs that do not reflect the underlying profile diversity. The paper concludes that prompt-level cognitive injection is insufficient and calls for below-the-prompt approaches such as fine-tuning and activation steering. All code, profiles, prompts, and model outputs are publicly released.
What's missing
The study does not address whether the Polymarket trader sample is representative of broader human forecasting diversity, or whether the null injection result might differ with more sophisticated prompt engineering beyond the tested approaches.
What different sources said
- arXiv cs.AICenter
Nous: An Attempt to Extract and Inject the Cognition Behind Prediction-Market Behavior
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.