Synthetic Players

Positioning

Related work — occupied territory and precise differentiation

What is already established, what collides, and the narrow triangle this paper defends.

The defensible novelty triangle

1. A registered strategic-interaction example where a fixed persona panel passes coarse marginal checks while continuation-probability estimates stay small and imprecise. 2. The mechanism-level pattern: dispersion carried largely by between-prompt composition of empirically corner-concentrated policies. 3. The credibility layer: registration, provenance, replay, mechanical adjudication, and public post-adjudication correction. Explicitly not claimed: first demonstration of realism/effect divergence, drift-free panels, human interiority, trait causality, or a p13 capability finding.

Closest occupied territory

WorkWhat it establishes
Li & Ji 2026When simulations look right but causal effects go wrong: LLMs as behavioral simulatorssource ↗
Ashokkumar et al. 2026Large language models can predict the results of social science experimentssource ↗
Persson et al. 2026Statistical foundations of LLM-based A/B testing: a surrogacy frameworksource ↗
Lin et al. 2026The illusion of intervention: your LLM-simulated experiment is an observational studysource ↗
Xie et al. 2026Evaluating the statistical realism of LLM-generated social science data (SSDataBench)source ↗
Harry et al. 2026Beyond fixed psychological personas: state beats trait, but language models are state-blindsource ↗
Xiao et al. 2026The chameleon's limit: persona collapse and homogenization in LLMssource ↗

Direct strategic-behavior collisions

WorkCollision
Akata et al. 2025Playing repeated games with large language modelssource ↗
Pal et al. 2026Strategies of cooperation and defection in five large language modelssource ↗
Georgousis et al. 2026Evaluating counterfactual strategic reasoning in large language modelssource ↗
Mousavi Davoudi et al. 2026Same game, different story: a strategic-robustness benchmarksource ↗
Mei et al. 2024A Turing test of whether AI chatbots are behaviorally similar to humanssource ↗

Synthetic participants and personas

WorkRelation
Bisbee et al. 2024Synthetic replacements for human survey data? The perils of large language modelssource ↗
Boelaert et al. 2025Machine bias: how do generative language models answer opinion polls?source ↗
Anthis et al. 2025Position: LLM social simulations are a promising research methodsource ↗
Hullman et al. 2026This human study did not involve human subjects: validating LLM simulationssource ↗
Park et al. 2024LLM agents grounded in self-reports enable general-purpose simulation of individualssource ↗
Argyle et al. 2023Out of one, many: using language models to simulate human samplessource ↗
Horton 2023Large language models as simulated economic agents (Homo Silicus)source ↗
Batzner et al. 2025Whose personae? Synthetic persona experiments and pathways to transparencysource ↗
Sclar et al. 2024Quantifying language models' sensitivity to spurious features in prompt designsource ↗
Shanahan et al. 2023Role play with large language modelssource ↗

Comparators and classical lineages

WorkUse here
Dal Bó & Fréchette 2011The evolution of cooperation in infinitely repeated gamessource ↗
Lucas 1976Econometric policy evaluation: a critiquesource ↗
Cronbach & Meehl 1955Construct validity in psychological testssource ↗
ICH E10 / Temple & Ellenberg 2000Assay sensitivity in controlled trialssource ↗
Windrum et al. 2007 / Grimm et al. 2005Agent-based model validation and equifinalitysource ↗
Statistical methodsStatistical foundations used by the analysessource ↗

Sources: the paper's §2 and References, plus the archived literature map and novelty-relationships documents (working research maps, not verdict-bearing).