arXiv subjects

Zhanzhan Zhao

Publications and source records attributed to Zhanzhan Zhao.

15 recordsLinked to original sources

arXiv2026

Mapping the Emerging Social Science of Large Language Models

As large language models (LLMs) blur the boundary between cognition and computation, they reshape knowledge production and collective life. We combine a curated review of 198 papers with computational mapping of 47,719 formally published papers from five databases. Text embeddings, clustering, and topic modeling organize the curated literature into 13 subcategories across three domains. LLM as Social Minds examines mind-like reasoning and socially grounded behavior in individual models. LLM Societies explores how interactions among agents produce collective behavior and social organization. LLM-Human Interactions examines how people perceive and engage with LLMs, and how interaction reshapes human thinking, creativity, and relationships. Structural topic modeling of the larger corpus traces how research attention shifts over time and differs across publication venues. We present a reproducible framework that identifies the field's underlying structure and enables its continuous refinement as new studies emerge. The review also identifies a gap in current understanding: the mechanisms underlying many of LLMs' observed cognitive and social behaviors remain unclear. Human-like performance can then be mistaken for understanding, obscuring the norms, biases, and structures of power that LLMs reproduce. We call for shared benchmarks and robustness checks across the three domains to build a reliable evidence base for understanding LLMs' social behaviors and informing their design and use.

cs.CY↗

arXiv2026

The Price of a Chance in Artificial Societies

Rapid advances in large language models (LLMs) have enabled autonomous agents to interact, coordinate, and pursue individual goals in artificial societies. However, our understanding of how reward structures shape their competitive behavior and collective outcomes remains incomplete. We simulate 200 LLM agents competing over 50 rounds to maximize individual wealth. We define overcompetition and lying flat as population investment above and below a Bayesian adaptive baseline, respectively. At the lowest tested reward, populations lie flat; at the two highest, they overcompete, most strongly at intermediate winner coverage. The investment committed to pursuing an uncertain reward constitutes the price of a chance. Agents near the qualification cutoff can invest heavily yet receive no prize, finishing poorer than nonwinners who invest less. Agents' stated rationales link investment decisions to perceived qualification prospects and expected returns. At the collective level, greater investment does not consistently improve ability selection. With a fixed total prize budget, offering smaller prizes to more winners can reduce investment while preserving selection quality relative to the corresponding top-ability target. Further broadening reward coverage yields additional savings but reduces selection quality. These findings show how reward design shapes overcompetition and lying flat, distributes the costs of pursuing opportunity, and influences what competition collectively achieves.

cs.CY↗

arXiv2026

Reversing Polarization through Tolerance-Aware Recommendation and Adaptive Diversification

Recommender systems face a tension between broadening exposure to diverse viewpoints and avoiding interactions that reinforce extreme positions. This tension is especially consequential in already polarized populations. Here we investigate whether recommendation strategies can reverse polarization rather than merely slow its development. Building on the Attraction--Repulsion Model of opinion dynamics, we introduce two complementary design principles: Tolerance-aware recommendation, which seeks to learn each individual's tolerance for disagreement and reduces exposure to viewpoints beyond that tolerance, and adaptive diversification, which broadens exposure for agents with more extreme opinions while retaining stronger personalization for moderates. Agent-based experiments show that highly personalized recommendations intensify polarization under polarized initial conditions. Tolerance awareness alone primarily slows polarization, whereas combining it with adaptive diversification can reduce polarization below its initial level and, under suitable conditions, restore consensus. Additional experiments incorporating persistent extreme influencers and a spiral-of-silence mechanism show that the combined strategy can mitigate polarization under social pressure and selective participation. To examine whether these findings extend beyond predefined opinion-update rules, we repeat the experiments on a simulated social platform populated by 100 LLM-driven agents prompted to move toward opinions similar to their own and away from those that differ substantially. Across three runs, the combined strategy is again the only recommendation policy tested that reverses polarization in every run. These findings distinguish polarization prevention from reversal and identify tolerance-aware, adaptive exposure as a promising recommendation design principle for reconnecting divided populations.

cs.SI↗

arXiv2026

Learning to Transfer School Attendance Zoning across Cities

Designing multi-school attendance zones requires selecting constraint settings that balance educational equity, accessibility, and school capacity. Repeating this process for each new city can require substantial local calibration and computational search. We introduce COMPASS, a parameter-to-policy framework that transfers knowledge from previous zoning searches to unseen cities through supervised prediction. Using search-derived parameter configurations as offline supervision, COMPASS learns to predict interpretable zoning-control parameters from urban representations and zoning-difficulty signals. A downstream mixed-integer linear program converts these parameters into explicit zoning plans, with conditional parameter repair when the initial solve is unsuccessful. We evaluate parameter prediction, zoning-outcome recovery, and computational efficiency through five-fold cross-validation across 3,000 US cities, complemented by a case study of Atlanta. COMPASS achieves a 99.80% solver-feasible success rate and a 49.14-fold median deployment-time speedup relative to repeated city-specific parameter search. Further analysis shows that transfer performance varies across urban contexts and policy dimensions: structurally difficult cities receive the largest feasibility gains, while enrollment balance remains the hardest outcome to recover. These findings demonstrate how cross-city learning can reduce repeated search while retaining explicit zoning controls and spatial plans for local inspection and adaptation.

cs.CY↗

arXiv2026

The Emergence of Social Science of Large Language Models

The social science of large language models (LLMs) examines how these systems evoke mind attributions, interact with one another, and transform human activity and institutions. We conducted a systematic review of 270 studies, combining text embeddings, unsupervised clustering and topic modeling to build a computational taxonomy. Three domains emerge organically across the reviewed literature. LLM as Social Minds examines whether and when models display behaviors that elicit attributions of cognition, morality and bias, while addressing challenges such as test leakage and surface cues. LLM Societies examines multi-agent settings where interaction protocols, architectures and mechanism design shape coordination, norms, institutions and collective epistemic processes. LLM-Human Interactions examines how LLMs reshape tasks, learning, trust, work and governance, and how risks arise at the human-AI interface. This taxonomy provides a reproducible map of a fragmented field, clarifies evidentiary standards across levels of analysis, and highlights opportunities for cumulative progress in the social science of artificial intelligence.

cs.AI↗

arXiv2026

Replay-Gated Neural Execution: Decoupling Persistent Behavioral Specifications from Neural Realizations in Frozen Language Models

Input-conditioned neural interventions raise a runtime question: what persists when one behavioral specification admits multiple actions whose validity depends on execution state? We introduce replay-gated neural execution, separating five objects: a persistent behavioral predicate, its state-indexed certified realization set, a transient action witness, a budget-limited finder, and execution authorization. Candidates undergo isolated FP32/BF16 replay of the frozen model; commitment additionally requires a valid run audit. Experiments on Qwen3-0.6B and SmolLM2-360M-Instruct establish distinct failure modes for these objects. Independent initializations yield distinct certified actions in all 24 tested fixed-state cells. Unchanged SmolLM2 witnesses remain certified in all 128 native states but only 66 of 384 off-diagonal transfers. All 767 archived Qwen witnesses replay successfully, yet a budget-limited finder misses one known-realizable cell in all three prespecified runs. Of 1,141 replay-submitted candidates, 174 fail item certification. A frozen three-tier cascade uses these boundaries to reject uncertified proposals and escalate audit-valid search misses. On 256 previously sealed Qwen Fresh requests, 221 first certify at the lowest-cost tier and all 256 receive audited authorization, with no observed bypass. Relative to frozen full search, the median singleton search-and-certification cost ratio is 0.1055 and P95 is 1.3485, including failed tiers. Within the studied behavioral family on two small models, these results support state-indexed, set-valued execution semantics: specifications persist, search proposes witnesses, and replay certification plus run audit grants execution authority.

cs.CL↗

arXiv2026

You Can't Fool Us: Understanding the Resilience of LLM-driven Agent Communities to Misinformation

Misinformation resilience is a dynamic community process: communities differ not only in whether they initially trust false claims, but also in how they recover through interaction, questioning, correction, and support withdrawal. We study this process with an LLM-based agent simulation that constructs synthetic communities along two theoretically motivated dimensions: Actively Open-minded Thinking (AOT), which captures evidence-seeking and willingness to revise beliefs, and Political Ideology (PI), which captures identity-based interpretation of contested claims. These two traits allow us to examine how evidence-oriented reasoning and ideological alignment jointly shape community responses to credible misinformation shocks. Across systematically varied AOT-PI communities, we find that higher AOT improves both resistance to misinformation uptake and recovery after trust peaks. PI shapes the recovery pathway: ideologically moderate communities recover more reliably, while polarized communities retain more residual support. Stance-level analysis shows that resilience depends on whether agents move from questioning a claim to denying or correcting it and withdrawing prior support. Intervention experiments further show that persuasion and fact checking better support post-peak correction, whereas accuracy prompts mainly induce early caution and source warnings have weaker effects. Together, this work provides a mechanism-level account of community misinformation resilience, showing how psychological composition and intervention design shape whether communities move from misinformation exposure toward correction or persistent support.

cs.CY↗

arXiv2026

Incentivizing User Data Contributions for LLM Improvement under Withdrawal Rights

The continued improvement of large language models (LLMs) increasingly depends on eliciting high-quality, user-generated data, yet such data are costly to provide and often withheld due to privacy and effort concerns. This creates a fundamental design challenge: how to incentivize data contribution when model improvements require coordinated, threshold-level inputs, while contributions remain privately costly and partially reversible. We develop and theoretically analyze incentive mechanisms for user data contribution that explicitly account for threshold effects and reversibility, focusing on how subsidies and withdrawal rights can be jointly designed to overcome coordination failure. As a natural benchmark, we first consider subsidy-based incentives, under which users respond to posted payments with privately optimal floor contributions. These decentralized responses may fall below the improvement threshold, resulting in subsidy expenditure without model improvements. We then analyze mechanisms with withdrawal rights, in which users report costs, the provider centrally assigns contribution burdens, and users may withdraw before training. We prove that combining cost reporting with personalized assignment can eliminate inefficient provision by ensuring that data are collected only when improvement is sustainable, converting infeasible instances into a null outcome rather than subsidy leakage. Finally, we compare two withdrawal protocols. The simultaneous protocol can achieve lower total cost, while the small-first sequential protocol better incentivizes participation, encouraging greater data provision and thereby increasing the probability of crossing the improvement threshold.

cs.GT↗

arXiv2026

Cities at Play: Improving Equilibria in Urban Neighbourhood Games

How should cities invest to improve social welfare when individuals respond strategically to local conditions? We model this question using a game-theoretic version of Schelling's bounded neighbourhood model, where agents choose neighbourhoods based on concave, non-monotonic utility functions reflecting local population. While naive improvements may worsen outcomes - analogous to Braess' paradox - we show that carefully designed, small-scale investments can reliably align individual incentives with societal goals. Specifically, modifying utilities at a total cost of at most $0.81 ε^2 \cdot \texttt{opt}$ guarantees that every resulting Nash equilibrium achieves a social welfare of at least $ε\cdot \texttt{opt}$, where $\texttt{opt}$ is the optimum social welfare. Our results formalise how targeted interventions can transform supra-negative outcomes into supra-positive returns, offering new insights into strategic urban planning and decentralised collective behaviour.

cs.GT↗

arXiv2025

The Emergence of Altruism in Large-Language-Model Agents Society

Leveraging Large Language Models (LLMs) for social simulation is a frontier in computational social science. Understanding the social logics these agents embody is critical to this attempt. However, existing research has primarily focused on cooperation in small-scale, task-oriented games, overlooking how altruism, which means sacrificing self-interest for collective benefit, emerges in large-scale agent societies. To address this gap, we introduce a Schelling-variant urban migration model that creates a social dilemma, compelling over 200 LLM agents to navigate an explicit conflict between egoistic (personal utility) and altruistic (system utility) goals. Our central finding is a fundamental difference in the social tendencies of LLMs. We identify two distinct archetypes: "Adaptive Egoists", which default to prioritizing self-interest but whose altruistic behaviors significantly increase under the influence of a social norm-setting message board; and "Altruistic Optimizers", which exhibit an inherent altruistic logic, consistently prioritizing collective benefit even at a direct cost to themselves. Furthermore, to qualitatively analyze the cognitive underpinnings of these decisions, we introduce a method inspired by Grounded Theory to systematically code agent reasoning. In summary, this research provides the first evidence of intrinsic heterogeneity in the egoistic and altruistic tendencies of different LLMs. We propose that for social simulation, model selection is not merely a matter of choosing reasoning capability, but of choosing an intrinsic social action logic. While "Adaptive Egoists" may offer a more suitable choice for simulating complex human societies, "Altruistic Optimizers" are better suited for modeling idealized pro-social actors or scenarios where collective welfare is the primary consideration.

cs.AI↗

arXiv2024

Modeling Processes of Neighborhood Change

An urban planner might design the spatial layout of transportation amenities so as to improve accessibility for underserved communities -- a fairness objective. However, implementing such a design might trigger processes of neighborhood change that change who benefits from these amenities in the long term. If so, has the planner really achieved their fairness objective? Can algorithmic decision-making anticipate second order effects? In this paper, we take a step in this direction by formulating processes of neighborhood change as instances of no-regret dynamics; a collective learning process in which a set of strategic agents rapidly reach a state of approximate equilibrium. We mathematize concepts of neighborhood change to model the incentive structures impacting individual dwelling-site decision-making. Our model accounts for affordability, access to relevant transit amenities, community ties, and site upkeep. We showcase our model with computational experiments that provide semi-quantitative insights on the spatial economics of neighborhood change, particularly on the influence of residential zoning policy and the placement of transit amenities.

cs.MA↗

arXiv2022

A Heterogeneous Schelling Model for Wealth Disparity and its Effect on Segregation

The Schelling model of segregation was introduced in economics to show how micro-motives can influence macro-behavior. Agents on a lattice have two colors and try to move to a different location if the number of their neighbors with a different color exceeds some threshold. Simulations reveal that even such mild local color preferences, or homophily, are sufficient to cause segregation. In this work, we propose a stochastic generalization of the Schelling model, based on both race and wealth, to understand how carefully architected placement of incentives, such as urban infrastructure, might affect segregation. In our model, each agent is assigned one of two colors along with a label, rich or poor. Further, we designate certain vertices on the lattice as "urban sites," providing civic infrastructure that most benefits the poorer population, thus incentivizing the occupation of such vertices by poor agents of either color. We look at the stationary distribution of a Markov process reflecting these preferences to understand the long-term effects. We prove that when incentives are large enough, we will have "urbanization of poverty," an observed effect whereby poor people tend to congregate on urban sites. Moreover, even when homophily preferences are very small, if the incentives are large and there is income inequality in the two-color classes, we can get racial segregation on urban sites but integration on non-urban sites. In contrast, we find overall mitigation of segregation when the urban sites are distributed throughout the lattice and the incentives for urban sites exceed the homophily biases. We prove that in this case, no matter how strong homophily preferences are, it will be exponentially unlikely that a configuration chosen from stationarity will have large, homogeneous clusters of agents of either color, suggesting we will have racial integration with high probability.

nlin.AO↗

arXiv2022

Mathematically Quantifying Non-responsiveness of the 2021 Georgia Congressional Districting Plan

To audit political district maps for partisan gerrymandering, one may determine a baseline for the expected distribution of partisan outcomes by sampling an ensemble of maps. One approach to sampling is to use redistricting policy as a guide to precisely codify preferences between maps. Such preferences give rise to a probability distribution on the space of redistricting plans, and Metropolis-Hastings methods allow one to sample ensembles of maps from the specified distribution. Although these approaches have nice theoretical properties and have successfully detected gerrymandering in legal settings, sampling from commonly-used policy-driven distributions is often computationally difficult. As of yet, there is no algorithm that can be used off-the-shelf for checking maps under generic redistricting criteria. In this work, we mitigate the computational challenges in a Metropolized-sampling technique through a parallel tempering method combined with ReCom[12] and, for the first time, validate that such techniques are effective on these problems at the scale of statewide precinct graphs for more policy informed measures. We develop these improvements through the first case study of district plans in Georgia. Our analysis projects that any election in Georgia will reliably elect 9 Republicans and 5 Democrats under the enacted plan. This result is largely fixed even as public opinion shifts toward either party and the partisan outcome of the enacted plan does not respond to the will of the people. Only 0.12% of the $\sim$160K plans in our ensemble were similarly non-responsive.

cs.CY↗

arXiv2022

Mitigating Filter Bubbles within Deep Recommender Systems

Recommender systems, which offer personalized suggestions to users, power many of today's social media, e-commerce and entertainment. However, these systems have been known to intellectually isolate users from a variety of perspectives, or cause filter bubbles. In our work, we characterize and mitigate this filter bubble effect. We do so by classifying various datapoints based on their user-item interaction history and calculating the influences of the classified categories on each other using the well known TracIn method. Finally, we mitigate this filter bubble effect without compromising accuracy by carefully retraining our recommender system.

cs.LG↗

arXiv2019

Alice's Adventures in the Markovian World

This paper proposes an algorithm Alice having no access to the physics law of the environment, which is actually linear with stochastic noise, and learns to make decisions directly online without a training phase or a stable policy as initial input. Neither estimating the system parameters nor the value functions online, the proposed algorithm generalizes one of the most fundamental online learning algorithms Follow-the-Leader into a linear Gauss-Markov process setting, with a regularization term similar to the momentum method in the gradient descent algorithm, and a feasible online constraint inspired by Lyapunov's Second Theorem. The proposed algorithm is considered as a mirror optimization to the model predictive control. Only knowing the state-action alignment relationship, with the ability to observe every state exactly, a no-regret proof of the algorithm without state noise is given. The analysis of the general linear system with stochastic noise is shown with a sufficient condition for the no-regret proof. The simulations compare the performance of Alice with another recent work and verify the great flexibility of Alice.

eess.SY↗