phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | LLM generative agents in the ultimatum game— “A multi-agent simulation, designed to improve the reasoning ability of generative agents through prompting methods, was developed to reproduce the result of an actual economic experiment on the ultimatum game.” | |
| claim-type | null-result— “Therefore, by improving the reasoning ability of generative agents, the result of LLMs-driven MAS does not align with the actual experimental result; rather, it closely aligns with the theoretical solution.” | |
| method — evidence | — “The simulation strategy involved generating 1000 independent decisions for each unique combination of role (proposer/responder), prompt type, and temperature.” | |
| transfer | from natural language processing / machine learning (prompt engineering): Prompt-engineering techniques for eliciting language-model reasoning are imported into the agent-based-modelling tradition of experimental economics, with prompts standing in for experimental instructions to human subjects.— “To situate our approach within existing methodological traditions, we note that our use of LLM-based generative agents as substitutes for human participants builds on the lineage of agent-based modeling in experimental economics, while integrating recent advances in natural language reasoning through prompt engineering.” | transfers[0] |
| mechanism | Persona settings act as a proxy that embeds social preferences and bounded rationality into the agents; without them the agents' improved reasoning drives them to the subgame-perfect solution instead of to human-like fairness behaviour— “This connection is further supported by our earlier observation that persona settings influenced the degree of deviation from human behavior: their inclusion may act as a proxy for embedding social preferences and bounded rationality into in LLM-driven MAS.” | mechanisms[0] |
| mechanism | Randomly generated persona attributes in earlier studies injected noise into agent behaviour, which masked the effect of reasoning ability and prevented it from being evaluated systematically— “The randomly generated persona attributes in previous studies likely acted as noise, limiting the systematic evaluation of reasoning-related outcomes.” | mechanisms[1] |
| mechanism | Skew in how the agents' offers were distributed across offer values drives the fitted acceptance-rate lines to impossible values above 1.0 or below 0, producing the anomalies seen in the responder-side figures— “This skewness accounts for the anomalies observed in the regression lines of Fig. 4, where acceptance rates exhibited extreme values, either exceeding 1.0 or falling below 0.” | mechanisms[2] |
| extends | Kitadai et al. (2023) LLM-driven ultimatum-game simulation— “The simulation framework in this study was constructed as an extension of the prompting structure introduced in Kitadai et al. (2023), incorporating methodological enhancements to examine the impact of reasoning ability in LLMs.” | claim_relations[0] |
| contradicts | Horton (2023) and Aher et al. (2023) reports of successful behavioural replication by LLM agents— “However, these successes may stem from coarse-grained evaluation criteria: both studies assessed behavior at the level of aggregate distributions or binary alignment with typical outcomes, without deeper analysis of underlying reasoning processes or individual-level deviations.” | claim_relations[1] |
| builds-on | prior LLM-agent studies that confounded reasoning with persona, payoff framing and temperature— “Our study addresses this gap by varying only the reasoning-related prompting strategy (zero-shot, few-shot, and chain-of-thought) while holding other factors constant.” | claim_relations[2] |
| sense | replication of human behaviour - AT AGGREGATE GRAIN: distributions of choices or binary alignment with typical outcomes, which is the criterion under which earlier work succeeded— “both studies assessed behavior at the level of aggregate distributions or binary alignment with typical outcomes” | senses[0] |
| sense | replication of human behaviour - AT INDIVIDUAL GRAIN: underlying reasoning processes and individual-level deviations, which the earlier evaluations did not examine— “without deeper analysis of underlying reasoning processes or individual-level deviations.” | senses[1] |
| sense | agent quality - split into two properties that trade off against each other: behavioral realism and rationality— “the results highlighted distinct trade-offs between behavioral realism and rationality” | senses[2] |
| sense | successful replication - in the prior literature a claim about AGGREGATE DISTRIBUTIONS or binary alignment with typical outcomes; this paper attributes those successes to the coarseness of the criterion rather than to the agents— “However, these successes may stem from coarse-grained evaluation criteria: both studies assessed behavior at the level of aggregate distributions or binary alignment with typical outcomes, without deeper analysis of underlying reasoning processes or individual-level deviations” | senses[3] |
| sense | One-shot ultimatum game - a demanding test that removes learning and reputation and supplies a unique equilibrium benchmark— “We focus on the one-shot ultimatum game for three main reasons. First, it eliminates history-dependent learning and reputation effects, leaving a unique subgame-perfect equilibrium that provides a clean theoretical benchmark. Second, recent cross-task evaluations by Mei et al. (2024), who compared large-language-model agents with human… | senses[4] |
| sense | Human low-offer rejection - payoff sacrifice explained as resistance to perceived unfairness— “Although s2 = accept is the weakly dominant strategy of player 2, in economic experiments it is common to observe decisions of s2 = reject when the value of s1 is low (Güth et al. 1982). A commonly proposed explanation for this phenomenon is that responders may choose to sacrifice their payoffs to resist perceived unfairness, particularly… | senses[5] |
| sense | Generative AI - distribution-learning synthesis rather than classification within existing boundaries— “Generative AI refers to a class of models that autonomously produce new, plausible data samples across various modalities—from text and images to audio and beyond. Unlike discriminative models that classify or predict within existing data boundaries, generative models learn the underlying data distribution itself, enabling them to synthesize… | senses[6] |
| sense | GPT-4 as behavioral model - a statistical knowledge base reflecting norms, heuristics, and communication conventions— “From a behavioral modeling standpoint, gpt-4 can be interpreted as a statistical knowledge base constructed from human language and interaction patterns. Its training corpus contains linguistic patterns that reflect social norms, strategic heuristics, and communicative conventions across cultural contexts (Lorè and Heydari 2024). This enables… | senses[7] |
| sense | Hallucination - plausible but wrong output arising from association-based rather than grounded modeling— “Nevertheless, several limitations must be addressed when deploying LLMs as behavioral agents. First, hallucinations, plausible but incorrect outputs, can emerge, especially under poorly constrained prompts (Bubeck et al. 2023). These are a byproduct of the model’s reliance on learned statistical associations rather than grounded world models.… | senses[8] |
| sense | Few-shot prompting - annotated examples guiding output style and logic— “While these limitations remain open challenges, recent work has shown that prompt design can shape the inferential structure of LLM outputs. We therefore adopt two prompting strategies—(i) few-shot prompting, where the model is shown a small number of annotated examples to guide output style and logic (Brown et al. 2020), and (ii)… | senses[9] |
| sense | Chain-of-thought prompting - explicit instruction to articulate intermediate reasoning steps | senses[10] |
| sense | Reasoning ability - observable coherent multi-step response structure, not a claim about mental models or logical soundness Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[11] |
| sense | LLMs as empirical tools - outputs analyzed, compared, and validated like participant data under safeguards Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[12] |
| sense | Replication success under coarse criteria - aggregate distributions or binary typical-outcome alignment without individual reasoning Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[13] |
| sense | Latent utility hypothesis - agents may optimize perceived social harmony rather than material self-interest Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[14] |
| sense | Isolated reasoning manipulation - vary zero-shot, few-shot, and chain-of-thought while holding persona, framing, and temperature outside the focal contrast Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[15] |
| sense | Core validity question - convergence toward normative rationality versus divergence from observed human behavior Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[16] |
| sense | Simulation sample - 1000 independent decisions balancing statistical resolution against chain-of-thought cost Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[17] |
| sense | Prompt engineering - methodological analog of instructions delivered to human experimental participants Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[18] |
| sense | AD versus KS - overall distributional difference versus maximum cumulative distribution difference Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[19] |
| sense | LLMs as human stand-ins - currently unwarranted without explicit social context and methodological scaffolding Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[20] |
| sense | Persona settings - configured cultural, age, and gender context needed because current models do not inherently possess it Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[21] |
| sense | Persona as behavioral proxy - possible embedding of social preferences and bounded rationality Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | senses[22] |
| control | the in-context examples RE-DRAWN AT RANDOM from the raw data and the whole simulation re-run several times, alongside runs at 15 and 20 examples — held — rules out the particular examples chosen for the prompt as the source of the result - an LLM's behaviour is conditioned on its examples, so a finding from one draw is a fact about that draw Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[0] |
| audit | reasoning ability —causes→ theoretical solution — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but higher reasoning moved results closer to the theoretical solution; causes is loose. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[0] |
| differs in | reasoning ability —causes→ theoretical solution — WHICH TARGET THE AGENTS CONVERGE ON, and the two targets pull apart: raising reasoning ability moves the simulation TOWARDS the game-theoretic solution and AWAY from the human experimental result. Better reasoning makes the agent a worse model of people - the paper's title asks the question and its abstract answers no Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[0] |
| weight | reasoning ability —causes→ theoretical solution — 0.9 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[0] |
| audit | reasoning ability —contrasted-with→ behavioral realism — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but higher reasoning moved results away from real experimental results; contrasted-with 'behavioral realism' is loose. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[2] |
| note | reasoning ability —contrasted-with→ behavioral realism — The record already carries a sense saying agent quality splits into two properties that TRADE OFF. This edge is that trade-off as the paper measured it. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[2] |
| weight | reasoning ability —contrasted-with→ behavioral realism — 0.8 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[2] |
| note | Generative AI —distinct-from→ discriminative models — The paper explicitly contrasts distribution-learning generation with classification and prediction. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[5] |
| note | opaque biases —causes→ harmful stereotypes — The paper says biases can perpetuate or amplify harmful stereotypes. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[6] |
| weight | opaque biases —causes→ harmful stereotypes — 0.7 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[6] |
| note | scaffolding methods —distinct-from→ general mitigation techniques — The authors explicitly adopt the strategies not as general mitigation but as reasoning scaffolds. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[7] |
| note | prompting strategies —causes→ coherence, inference depth, and contextual relevance — Prior work is reported to show these strategies enhance all three output properties. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[8] |
| weight | prompting strategies —causes→ coherence, inference depth, and contextual relevance — 0.9 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[8] |
| audit | perceived social harmony —contrasted-with→ material self-interest — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the claim is what Mei et al. (2024) 'proposed'; attributed. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[9] |
| note | perceived social harmony —contrasted-with→ material self-interest — The proposed implicit objective is social harmony rather than material self-interest. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[9] |
| weight | perceived social harmony —contrasted-with→ material self-interest — 0.7 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[9] |
| note | treating LLMs as direct substitutes for human subjects —is-a→ flawed approach — The paper directly classifies treating LLMs as substitutes as a flawed approach. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[10] |
| note | contextual variables —causes→ human decision-making — Culture, age, and gender are said to significantly influence human decision-making. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[11] |
| note | Variability in reasoning ability and persona configurations —causes→ agent behavior — The paper states this variability can significantly influence agent behavior. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[12] |
| weight | Variability in reasoning ability and persona configurations —causes→ agent behavior — 0.7 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[12] |
| outcome (stated by the tagger, no quotation) | control — held | controls[0] |
| for which use | limitation — the transfer to agents with personas: WHILE OUR METHODOLOGICAL CHOICE CLARIFIES THE ROLE OF REASONING ABILITY, IT ALSO REPRESENTS A LIMITATION OF THIS STUDY - the excluded factor is known to matter Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[0] |
| for which use | limitation — the precision of every estimate: WHILE THIS NUMBER IS NOT DERIVED FROM A FORMAL POWER ANALYSIS, IT WAS SELECTED BASED ON TWO MAIN CONSIDERATIONS Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[1] |
| for which use | limitation — the scope of the conclusions: BROADER APPLICATIONS TO MARKET-LEVEL MODELING OR INSTITUTIONAL DESIGN REMAIN IMPORTANT DIRECTIONS FOR FUTURE WORK Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[2] |
| for which use | limitation — published findings that generative agents reproduce human behaviour in economic games: this paper states those successes may stem from coarse-grained criteria — aggregate distributions or binary alignment — without individual-level or reasoning-process analysis Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[3] |
| for which use | limitation — the assumption that a more capable model will simulate an economic experiment better: this paper reports zero-shot prompting with an older model aligning closely with human behaviour, and states improving reasoning ability cannot lead directly to reproduction Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[4] |
| weight | LLM agents → traditional economic experiments — 0.7 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[0] |
| weight | Persona settings → social preferences — 0.6 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[6] |
| weight | Persona settings → bounded rationality — 0.6 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[7] |
| replication | untested | |
| no type for this | LLM agents → traditional economic experiments — LLM agents are proposed to augment rather than replace traditional economic experiments. Augments-but-does-not-replace is absent from the eight types; distinct-from would lose the methodological complement relation. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[0] |
| no type for this | ground-truth benchmarks → LLM-based simulations — Human data remain indispensable benchmarks for calibrating LLM simulations and identifying their boundary conditions. Benchmark-for/calibrates is absent from the eight types and is not mere operationalisation. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[1] |
| no type for this | outputs of generative agents → simulation design choices — Generative-agent outputs are highly sensitive to simulation design choices. Depends-on/sensitive-to is absent; causation would overstate a design dependence with multiple interacting parameters. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[2] |
| no type for this | foundation models → downstream tasks — Foundation models are adaptable to downstream tasks without retraining. Adaptable-to is absent from the eight types; neither kindhood nor operationalisation captures transfer to a task. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[3] |
| no type for this | prompt engineering → experimental instructions — Prompt engineering serves as an analog to instructions given to human participants. Analogous-to is absent; same-as would erase the distinction between machine prompts and human-study instructions. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[4] |
| no type for this | social context → stand-ins for real human groups — Prompts incorporating social context are a necessary condition for advanced LLMs to serve as human-like stand-ins. Prerequisite-for is absent from the eight relation types. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[5] |
| no type for this | Persona settings → social preferences — Persona inclusion may act as a proxy for embedding social preferences in the simulation. Proxy-for is absent; operationalised-as would claim a measurement definition the paper does not make. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[6] |
| no type for this | Persona settings → bounded rationality — Persona inclusion may proxy bounded rationality in LLM-driven simulations. Proxy-for is absent from the eight types and the paper presents this as a hypothesis. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[7] |
| limitation | Excluding persona settings in order to isolate reasoning ability is itself a limitation, since personas are known to shape human-like behaviour. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[0] |
| limitation | The sample size of 1000 decisions per condition was chosen pragmatically rather than by a formal power analysis. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[1] |
| limitation | Scope is restricted to decision-making and negotiation tasks; market-level and institutional applications are left to future work. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[2] |
| limitation | Earlier claims of success are attributed to the coarseness of the evaluation rather than to the agents: aggregate distributions and binary alignment, with no individual-level analysis. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[3] |
| limitation | The expectation the title asks about is not borne out: greater reasoning ability does not directly produce reproduction of the experiments. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[4] |
| limitation | The paper's own explanation of why earlier work succeeded is a claim about MEASUREMENT RESOLUTION, not about behaviour: prior studies assessed aggregate distributions or binary alignment 'without deeper analysis of underlying reasoning processes or individual-level deviations'. On that account the earlier results are not wrong, they are coarse - which is a harder thing to test. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[5] |
| limitation | Agent outputs are reported as highly sensitive to design choices - persona, prompt formulation, role instructions - so any single configuration's result is conditional on choices the paper had to make and could not vary at once. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[6] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | Raising the reasoning ability of LLM-driven agents pushed their ultimatum-game behaviour toward the game-theoretic solution and away from the real human experimental data. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[0] |
| finding | No prompting condition produced a distribution that statistically matched either the theoretical solution or the human experimental data. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[1] |
| finding | On the proposer side the older, weaker-reasoning model with zero-shot prompting was also relatively close to human behaviour, showing that better reasoning does not straightforwardly buy better reproduction of economic experiments. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[2] |
| finding | On the responder side, chain-of-thought prompting produced the closest match to the theoretical (rational) distribution of any condition. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[3] |
| finding | Against human responder data the ranking inverted: the weakest-reasoning condition, zero-shot gpt-3.5, matched human acceptance behaviour most closely. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[4] |
| finding | The authors conclude that LLM agents do not spontaneously behave like humans in economic games without prompts carrying social context, so using LLM-driven multi-agent simulations as substitutes for real human groups is currently unwarranted. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[5] |
| finding | At every temperature setting almost all agents offered a 50-50 split of coins. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[6] |
| finding | Because every other setting was identical to the earlier study, the authors attribute the difference in results to the presence or absence of persona settings, and conclude that personas are essential for applying LLM-driven simulation to economic experiments. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[7] |
| finding | Robustness checks with 15 and 20 in-prompt examples changed the synthesised distributions only slightly and left the comparative pattern across conditions unchanged. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[8] |
| quantity | 1000 — independent agent decisions generated for every combination of role, prompt type and temperature Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[0] |
| quantity | 10 — real experimental data points used as in-prompt examples for few-shot and chain-of-thought prompting Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[1] |
| quantity | KS = 0.479 and AD = 128 — smallest divergence from the human proposer distribution, obtained by the chain-of-thought condition Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[2] |
| quantity | AD = 15.5 — chain-of-thought responder distribution's divergence from the theoretical distribution, the smallest of any condition Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[3] |
| quantity | AD = 20.8 ... AD = 43.7 — divergence from human responder data for few-shot (third best) and for chain-of-thought (worst of the four conditions) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[4] |
| quantity | approximately 0.7 / approximately 0.4 or less — human acceptance rate at an offer of 50 versus the simulated agents' acceptance rate at the same offer Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[5] |
| variable | reasoning-related prompting strategy (independent) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.variables[0] |
| variable | temperature parameter (independent) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.variables[1] |
| variable | GPT model (independent) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.variables[2] |
| variable | acceptance rate (dependent) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.variables[3] |
| variable | persona settings (control) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.variables[4] |
| definition | reasoning ability Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.definitions[0] |
| definition | generative agent Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.definitions[1] |
| definition | Platonic Representation Hypothesis Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.definitions[2] |
| concept | generative agent Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[0] |
| concept | economic experiments Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[1] |
| concept | behavioral realism Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[2] |
| concept | reasoning ability Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[3] |
| concept | theoretical solution Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[4] |
| concept | persona Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[5] |
| concept | one-shot ultimatum game — not defined in this paper's own words; single-round bargaining design chosen to remove learning and reputation while preserving a demanding human-AI gap Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[6] |
| concept | history-dependent learning — not defined in this paper's own words; cross-round adaptation eliminated by the single-round design Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[7] |
| concept | reputation effects — not defined in this paper's own words; cross-round social consequence eliminated by the single-round design Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[8] |
| concept | subgame-perfect equilibrium — not defined in this paper's own words; unique game-theoretic solution used as a clean benchmark Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[9] |
| concept | theoretical benchmark — not defined in this paper's own words; equilibrium reference against which human-like bargaining is compared Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[10] |
| concept | human-AI performance gap — not defined in this paper's own words; performance discrepancy reported as largest for the selected game Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[11] |
| concept | stringent test bed — not defined in this paper's own words; interpretable and demanding environment for testing enhanced reasoning Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[12] |
| concept | artificial correlations — not defined in this paper's own words; cross-round dependence that chain-of-thought leakage could induce Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[13] |
| concept | weakly dominant strategy — not defined in this paper's own words; acceptance strategy that maximizes payoff weakly under the formal game Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[14] |
| concept | perceived unfairness — not defined in this paper's own words; social interpretation motivating costly rejection of low offers Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[15] |
| concept | LLM agents — not defined in this paper's own words; language-model-driven agents proposed as an augmenting experimental tool Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[16] |
| concept | traditional economic experiments — not defined in this paper's own words; human-subject economic studies the agents are intended to augment rather than replace Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[17] |
| concept | additional methodological tool — not defined in this paper's own words; supplementary role proposed for LLM agents Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[18] |
| concept | opaque biases — not defined in this paper's own words; training- and development-rooted distortions in behavioral output Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[19] |
| concept | harmful stereotypes — not defined in this paper's own words; social harms that model bias can perpetuate or amplify Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[20] |
| concept | structural inequities — not defined in this paper's own words; systemic harms that model bias can perpetuate or amplify Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[21] |
| concept | ground-truth benchmarks — not defined in this paper's own words; real human data used to calibrate and bound LLM simulations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[22] |
| concept | LLM-based simulations — not defined in this paper's own words; agent simulations calibrated against human population behavior Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[23] |
| concept | boundary conditions — not defined in this paper's own words; conditions separating faithful approximation from systematic deviation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[24] |
| concept | outputs of generative agents — not defined in this paper's own words; simulated decisions sensitive to persona, prompt, and role design Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[25] |
| concept | simulation design choices — not defined in this paper's own words; persona, prompt, and role settings on which outputs depend Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[26] |
| concept | Generative AI — not defined in this paper's own words; model class producing plausible new samples across modalities Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[27] |
| concept | generative models — not defined in this paper's own words; models learning a data distribution to synthesize new outputs Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[28] |
| concept | discriminative models — not defined in this paper's own words; models classifying or predicting within existing data boundaries Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[29] |
| concept | underlying data distribution — not defined in this paper's own words; distribution learned by generative rather than discriminative models Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[30] |
| concept | recognition to generation — not defined in this paper's own words; paradigm shift the authors connect to social modeling Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[31] |
| concept | methodological transformation — not defined in this paper's own words; change in the modeling of social phenomena enabled by generation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[32] |
| concept | LLMs — not defined in this paper's own words; GPT-family subclass of generative AI built on Transformers Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[33] |
| concept | key subclass — not defined in this paper's own words; taxonomic status assigned to LLMs within generative AI Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[34] |
| concept | Transformer architecture — not defined in this paper's own words; neural architecture underlying the GPT-family models Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[35] |
| concept | statistical knowledge base — not defined in this paper's own words; behavioral interpretation of GPT-4 as patterns learned from language and interaction Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[36] |
| concept | social norms — not defined in this paper's own words; human interaction regularities reflected in the training corpus Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[37] |
| concept | strategic heuristics — not defined in this paper's own words; decision patterns reflected in the training corpus Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[38] |
| concept | foundation models — not defined in this paper's own words; general-purpose model substrates adaptable across tasks Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[39] |
| concept | general-purpose substrates — not defined in this paper's own words; general-purpose characterization of foundation models Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[40] |
| concept | downstream tasks — not defined in this paper's own words; new uses reached through adaptation without retraining Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[41] |
| concept | retraining — not defined in this paper's own words; model rebuilding said not to be required for downstream adaptation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[42] |
| concept | hallucinations — not defined in this paper's own words; plausible but incorrect outputs under weak prompt constraints Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[43] |
| concept | learned statistical associations — not defined in this paper's own words; ungrounded association mechanism offered for hallucination Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[44] |
| concept | grounded world models — not defined in this paper's own words; world representations contrasted with learned statistical associations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[45] |
| concept | black boxes — not defined in this paper's own words; interpretability limitation of behavioral LLM agents Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[46] |
| concept | interpretability — not defined in this paper's own words; access limitation: outputs are analyzable but internal reasoning is not Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[47] |
| concept | prompt design — not defined in this paper's own words; control used to shape output inferential structure Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[48] |
| concept | few-shot prompting — not defined in this paper's own words; prompting with a small set of annotated demonstrations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[49] |
| concept | chain-of-thought (CoT) prompting — not defined in this paper's own words; prompting that elicits intermediate reasoning steps Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[50] |
| concept | general mitigation techniques — not defined in this paper's own words; broad correction role explicitly denied for the two strategies Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[51] |
| concept | scaffolding methods — not defined in this paper's own words; role assigned to prompting strategies for structured decisions Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[52] |
| concept | prompting strategies — not defined in this paper's own words; few-shot and chain-of-thought methods considered together Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[53] |
| concept | reasoning quality — not defined in this paper's own words; quality of agent decision reasoning influenced by scaffolding Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[54] |
| concept | coherence, inference depth, and contextual relevance — not defined in this paper's own words; three output properties reported as enhanced by prompting strategies Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[55] |
| concept | contextually coherent, multi-step responses — not defined in this paper's own words; observable response form used in the paper's reasoning definition Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[56] |
| concept | inferential structure — not defined in this paper's own words; pattern in generated responses rather than an asserted mental model Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[57] |
| concept | observable behavioral property — not defined in this paper's own words; behavioral rather than mentalistic status of reasoning ability Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[58] |
| concept | experimental variable — not defined in this paper's own words; status assigned to reasoning ability in the simulation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[59] |
| concept | pragmatist perspective — not defined in this paper's own words; stance evaluating useful behavioral approximation without claiming cognition Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[60] |
| concept | human cognition — not defined in this paper's own words; target the authors explicitly do not claim GPT-4 replicates Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[61] |
| concept | behavioral patterns — not defined in this paper's own words; economic-experiment regularities LLMs may approximate Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[62] |
| concept | methodological tools — not defined in this paper's own words; specific-context role justified by adequate behavioral fidelity Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[63] |
| concept | well-defined simulation contexts — not defined in this paper's own words; narrow settings in which methodological use may be justified Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[64] |
| concept | empirical tools — not defined in this paper's own words; LLMs treated as output-producing entities subject to comparison and validation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[65] |
| concept | evaluation protocols — not defined in this paper's own words; safeguard required for interpretable behavioral simulations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[66] |
| concept | methodological safeguards — not defined in this paper's own words; controls required alongside careful prompting and evaluation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[67] |
| concept | coarse-grained evaluation criteria — not defined in this paper's own words; aggregate or binary evaluation that may explain earlier replication successes Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[68] |
| concept | aggregate distributions — not defined in this paper's own words; population-level behavior grain used by prior evaluations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[69] |
| concept | individual-level deviations — not defined in this paper's own words; fine-grained differences omitted by aggregate evaluations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[70] |
| concept | finer-grained behavioral criteria — not defined in this paper's own words; stricter evaluation grain linked to failures of alignment Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[71] |
| concept | game-theoretic environments — not defined in this paper's own words; more complex strategic settings used in stricter evaluations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[72] |
| concept | failures of alignment — not defined in this paper's own words; systematic failures highlighted under finer evaluation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[73] |
| concept | behavioral Turing tests — not defined in this paper's own words; tests formalizing systematic social tendencies in LLM agents Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[74] |
| concept | latent utility function — not defined in this paper's own words; hypothesized implicit objective behind prosocial or cooperative behavior Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[75] |
| concept | perceived social harmony — not defined in this paper's own words; hypothesized implicit optimization target Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[76] |
| concept | material self-interest — not defined in this paper's own words; game-theoretic motive contrasted with social harmony Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[77] |
| concept | Persona design — not defined in this paper's own words; assignment of demographic or personality traits to agents Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[78] |
| concept | behavioral tendencies — not defined in this paper's own words; agent dispositions nudged by specified traits Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[79] |
| concept | causal effects — not defined in this paper's own words; effects obscured when several design parameters vary together Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[80] |
| concept | reasoning-related prompting strategy — not defined in this paper's own words; zero-shot, few-shot, and chain-of-thought manipulation varied in isolation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[81] |
| concept | confounding variables — not defined in this paper's own words; persona, framing, and temperature factors held outside the focal manipulation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[82] |
| concept | central research question — not defined in this paper's own words; whether better reasoning improves human alignment or amplifies rational deviation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[83] |
| concept | normative benchmarks of rationality — not defined in this paper's own words; game-theoretic rationality target toward which sophistication may converge Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[84] |
| concept | empirically grounded human behavior — not defined in this paper's own words; descriptive human target from which sophisticated agents may diverge Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[85] |
| concept | simulation framework — not defined in this paper's own words; three-step prompting structure extending earlier work Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[86] |
| concept | three sequential steps — not defined in this paper's own words; game explanation, decision instructions, and output specification Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[87] |
| concept | output format — not defined in this paper's own words; third stage constraining the agent response Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[88] |
| concept | 1000 independent decisions — not defined in this paper's own words; per-condition sample generated for each role, prompt type, and temperature Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[89] |
| concept | formal power analysis — not defined in this paper's own words; sample-size derivation explicitly not used Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[90] |
| concept | robustness check — not defined in this paper's own words; repeat and expanded-example analyses supporting result stability Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[91] |
| concept | computational cost per generation — not defined in this paper's own words; chain-of-thought expense limiting feasible sample size Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[92] |
| concept | statistical resolution — not defined in this paper's own words; precision side of the chosen sample-size trade-off Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[93] |
| concept | feasibility in simulation-based research — not defined in this paper's own words; resource side of the sample-size trade-off Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[94] |
| concept | substitutes for human participants — not defined in this paper's own words; methodological role located in agent-based modeling lineage Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[95] |
| concept | agent-based modeling — not defined in this paper's own words; experimental-economics tradition extended by LLM agents Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[96] |
| concept | natural language reasoning — not defined in this paper's own words; recent capability integrated through prompting Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[97] |
| concept | prompt engineering — not defined in this paper's own words; LLM instruction design treated as analogous to experimental instructions Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[98] |
| concept | Monte Carlo methods — not defined in this paper's own words; repeated-sampling tradition paralleled by the simulation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[99] |
| concept | experimental instructions — not defined in this paper's own words; human-study counterpart to prompt engineering Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[100] |
| concept | human-subject experiments — not defined in this paper's own words; methodological consistency target for the prompt structure Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[101] |
| concept | Anderson-Darling (AD) — not defined in this paper's own words; distribution-comparison test used alongside KS Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[102] |
| concept | Kolmogorov-Smirnov (KS) — not defined in this paper's own words; maximum cumulative-difference test used alongside AD Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[103] |
| concept | overall distributional differences — not defined in this paper's own words; overall discrepancy evaluated by the two tests Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[104] |
| concept | maximum cumulative distribution differences — not defined in this paper's own words; largest cumulative discrepancy evaluated by KS Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[105] |
| concept | social context — not defined in this paper's own words; prompt content required for advanced agents to approach human-like economic behavior Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[106] |
| concept | stand-ins for real human groups — not defined in this paper's own words; role the paper currently judges unwarranted Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[107] |
| concept | treating LLMs as direct substitutes for human subjects — not defined in this paper's own words; replacement of human subjects explicitly rejected Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[108] |
| concept | flawed approach — not defined in this paper's own words; paper's classification of direct LLM substitution Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[109] |
| concept | methodological scaffolding — not defined in this paper's own words; deliberate structure required for effective economic modeling Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[110] |
| concept | Persona settings — not defined in this paper's own words; explicitly configured context shaping human-like simulated behavior Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[111] |
| concept | contextual variables — not defined in this paper's own words; culture, age, and gender factors encoded through personas Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[112] |
| concept | human decision-making — not defined in this paper's own words; behavior significantly influenced by contextual variables Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[113] |
| concept | inequity aversion — not defined in this paper's own words; behavioral-economics account of low-offer rejection Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[114] |
| concept | fairness equilibrium — not defined in this paper's own words; alternative fairness-based account of ultimatum behavior Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[115] |
| concept | bounded rationality — not defined in this paper's own words; account of human departures from subgame-perfect equilibrium Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[116] |
| concept | social preferences — not defined in this paper's own words; human motive persona inclusion may proxy in LLM simulations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[117] |
| concept | Variability in reasoning ability and persona configurations — not defined in this paper's own words; combined design variability affecting agent behavior Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[118] |
| concept | agent behavior — not defined in this paper's own words; simulation output sensitive to reasoning and persona configurations Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[119] |
| concept | simulation outcomes — not defined in this paper's own words; outputs whose realism, interpretability, and applicability require design care Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[120] |
| Anderson-Darling test AD | as written: Anderson-Darling (AD) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.statistics[0] |
| Kolmogorov-Smirnov test KS | as written: Kolmogorov-Smirnov (KS) Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.statistics[1] |
| p-value p | Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.statistics[2] |
| regression | as written: piecewise linear regression Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.statistics[3] |
| confidence interval | Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.statistics[4] |
When we have changed our mind. Some entries in the table begin with a word in capitals and the reason for it. EVIDENCE CORRECTED — the finding still stands, but we had quoted the wrong sentence from the paper, and the quotation has been replaced with the right one. RETRACTED — the entry itself was wrong; it stays on the page, marked, with the reason, rather than being deleted. WITHDRAWN AND REPLACED — the entry was wrong, and a corrected version of it appears in the entries below. Each mark carries its own reason, in brackets, because the reasons differ from one entry to the next. An entry we have WITHDRAWN is removed from this page and kept in the record, with its reason. A line reading Audit and a date is a later re-reading of the entry against its whole sentence: DOUBTFUL means it was kept, and the reason says what is loose about it. Quotation marks inside the tagger's sentence. Where a tag is the tagger's own sentence, a phrase in quotation marks inside it is borrowed from the paper — the tagger's wording around it is not. The paper's own sentence is the evidence quote for that row; on this page it is withheld with the article, and the article is at its publisher. Nothing inside quotation marks anywhere on this page is ours: if it is quoted, it is the paper's. Whose words a concept carries. A concept entry records a term the paper itself uses, so the term, the sense given for it, and its evidence are the paper's own words, quoted where the paper fixes them. This is the opposite of a sense entry in the tag table, which is the tagger's sentence about how the paper uses a term. Where a concept entry could not be grounded in the paper's words, the entry says so on its own row.
Tag layer © the Meta-Tagging Project, CC BY 4.0. The paper text is © its authors/publisher
and is linked, not reproduced. Excerpts shown are short quotations for identification and verification.
· back to index
· concepts · authors
· עברית
© 2026 Shir Sivroni · ORCID 0009-0000-2597-2824