The DOI matches its Crossref record; 99 of the 101 references it cites were found in a registry or an index, and 1 are not checked yet; 13 of the 23 sources its claims cite were checked for support.
Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent Artificial Intelligence Systems
Open accessAbstract
Abstract Modern engineered systems increasingly involve complex sociotechnical environments where multiple agents—including humans and the emerging paradigm of agentic AI powered by large language models (LLMs)—must navigate social dilemmas that pit individual interests against collective welfare. As engineered systems evolve toward multi-agent architectures with autonomous LLM-based agents, traditional governance approaches using static rules or fixed network structures fail to address the dynamic uncertainties inherent in real-world operations. This article presents a novel framework that integrates adaptive governance mechanisms directly into the design of sociotechnical systems through a unique separation of agent interaction networks from information flow networks. We introduce a system comprising strategic LLM-based system agents that engage in repeated interactions and a reinforcement learning (RL)-based governing agent that dynamically modulates information transparency. Unlike conventional approaches that require direct structural interventions or payoff modifications, our framework preserves agent autonomy while promoting cooperation through adaptive information governance. The governing agent learns to strategically adjust information disclosure at each time-step, determining what contextual or historical information each system agent can access. Experimental results demonstrate that this RL-based governance significantly enhances cooperation compared to static information-sharing baselines. This work establishes information transparency as a dynamic design parameter and demonstrates how governance considerations can be effectively embedded into complex engineering systems from the design phase. While validated on the repeated Prisoner’s dilemma, which represents a challenging governance problem in an abstract model, our framework offers a flexible strategy for fostering desired collective outcomes across diverse sociotechnical engineering applications, from human–robot collaboration to autonomous vehicle networks and future multi-agent AI systems.
Verified 2026-10-05 · 1 reference not checked yet
Overview
CoChat Summary
TL;DR
The paper proposes adaptive information modulation as a governance mechanism for multi-agent AI systems, separating a fixed interaction network from an RL-controlled information network. In repeated Prisoner’s Dilemma experiments with LLaMa3-70b agents, an RL manager learned dynamic, heterogeneous information-disclosure policies that improved cooperation and social welfare over static baselines.
Key findings
- The RL manager outperformed all static baselines (LA, LA+NR, LA+AR) on normalized social welfare and cooperation rate across 50 runs; RL cooperation reached 100% [C,C] by the 10th timestep, while LA and LA+NR increased more slowly and did not converge to 100%, and LA+AR drove the system toward defection.
- Micro-level validation showed LLaMa3-70B followed a win-stay, lose-shift pattern: after mutual cooperation, cooperation was 87% when the coplayer was described as cooperating “Sometimes” and 99% when “Often”; after mutual defection with a coplayer described as cooperating “Often,” cooperation occurred 68% of the time.
- The learned policy used LA more in early steps 2–6, with about 20% mixed interventions, then consistently selected neighborhood ratio (LA+NR) from step 7 onward. High-degree nodes (mean 5.68) were significantly more likely to receive LA+NR than low-degree nodes (mean 4.64), p < 0.001, Cohen’s d = 0.60; agents receiving LA+NR had higher prior cooperation (mean 0.908) than those receiving LA (mean 0.801), p < 0.001, Cohen’s d = 2.61.
- Quantitative cooperation rates were unreliable: with opponent ratio fixed at 75%, an own ratio of 69.99% yielded 1% cooperation but 70% yielded 96%; typos in prompts also substantially shifted LLM behavior.
Methods
- Framework: 20 LLM agents on a fixed Erdős–Rényi network (link probability 0.25) per round, 20 timesteps, repeated Prisoner’s Dilemma (CC = 3 each; CD/DC = 5/0; DD = 1 each); 50 independently generated network topologies; LLaMa3-70b via LangChain and Groq, temperature 0.8.
- RL manager: Actor-Critic POMDP that observes aggregated states and selects one of three information types per agent per interaction—LA, LA+AR, or LA+NR—with one hidden layer of 256 neurons, actor learning rate 0.001, critic learning rate 0.005, and discount factor 0.99; trained offline to maximize discounted social welfare.
- Analysis: micro-level validation with N typically greater than 100 prompt repetitions and qualitative cooperation categories (Rarely <33%, Sometimes 33–66%, Often >66%); RL compared against static baselines; t-tests for degree centrality and prior cooperation differences.
Limitations
- Stated: limited evaluation sample (50 network instances, 20 timesteps) due to computational constraints; validation only on Prisoner’s Dilemma in random networks; LLM behavioral fidelity to humans requires controlled human-subject validation; coarse three discrete information strategies; assumes complete observability; field deployment remains future work.
- Stated: future work should test other network structures and games, finer or adaptive information granularity, partial observability, and integration with other governance mechanisms.
- Evident: micro-validation used only LLaMa3-70b and found sensitivity to typos and quantitative thresholds (e.g., 69.99% vs. 70%), raising robustness and generalizability concerns across LLMs and prompt variations.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
8 Valid · 4 Not valid · 8 Not checked · 125 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Recent findings indicate that the primary bottlenecks in multi-LLM systems arise not from individual model limitations, but from poor coordination and systemic misalignment [67].”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 2.3. LLMs in Multi-Agent Systems
What the sources say
- Supportedcites [67] Cemri, 2025 source
Section 5.3 says many multi-agent system failures arise from organizational design and agent coordination rather than limitations of individual agents. Section 4 likewise identifies system-design issues and inter-agent misalignment as failure categories. Page numbers are not provided in our copy. (LLM, conf=0.88)
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Still, foundational limitations (such as hallucination, autoregressive instability, and scaling bottlenecks) continue to constrain the reliability of large-scale, LLM-mediated multi-agent systems [77].”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 2.3. LLMs in Multi-Agent Systems
What the sources say
- Partly supportedcites [77] Tran, 2025 sourceExperimental — accuracy not yet measured
The Introduction (section 1) lists hallucination, autoregressive nature, and scaling laws as intrinsic LLM limitations. Section 2.2 describes cascading hallucinations in multi-agent interactions, while section 6.3 says those risks can scale and compromise reliability; the scalability discussion also identifies bottlenecks as agent populations grow. The copy does not specifically describe autoregressive instability or establish that autoregression itself constrains multi-agent reliability. (LLM, conf=0.84)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The influence of information on agent behavior is well-established through two mechanisms.First, agents make decisions under uncertainty, and information affects belief updates via Bayesian updating [32].Second, information enables social learning, where agents adopt observed successful strategies [28,29].”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > Why Information Modulation as a Governance Mechanism?
What the sources say
- Partly supportedcites [32] Acemoglu, 2011, Opinion Dynamics and Learning in Social Networks, Dyn. Games Appl. sourceExperimental — accuracy not yet measured
Section 2 (p. 9) describes agents using new information to update beliefs, including through Bayes’ rule and observations of others’ actions. The social-learning discussion (p. 24) likewise describes agents observing others’ decisions, but it does not say they adopt strategies because those strategies were observed to succeed. (LLM, conf=0.82)
- Partly supportedcites [28] Stewart, 2016, Small Groups and Long Memories Promote Cooperation, Sci. Rep. sourceExperimental — accuracy not yet measured
The Abstract (page unavailable) says individuals update behavior in light of past experience, supporting the broad point that information can affect behavior. But the paper models memory-based strategies in repeated games and, in Methods (page unavailable), payoff-based reproduction in a population; it does not establish that agents socially learn by adopting observed successful strategies. The Bayesian-updating statement carries a separate citation [32]. (LLM, conf=0.79)
- Could not verifycites [29] Bradley, 2018, Does Observability Affect Prosociality?, Proc. R. Soc. B: Biological Sci. sourceExperimental — accuracy not yet measured
The citation is attached to the social-learning claim. The Discussion and Introduction > Theoretical moderators describe how being observed can affect prosocial behavior through reputation and reciprocity, but they do not address agents adopting observed successful strategies. Searches for social learning and copying successful strategies found no passage supporting that mechanism; no page numbers are available in our copy. (LLM, conf=0.73)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“The RL manager serves as the governing entity within our framework and is trained using deep reinforcement learning, essentially using a two-layer governance scheme similar to [22,78].”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 3.4. RL Manager
What the sources say
- Not checked — no full text of the sourcecites [22] Chen, 2022, Dynamic Resource Allocation in Systems-of-Systems Using a Heuristic-Based Interpretable Deep Reinforcement Learning, ASME J. Mech. Des. source
- Supportedcites [78] Chen, 2024, The Sos Conductor: Orchestrating Resources With Iterative Agent-Based Reinforcement Learning, Syst. Eng. source
Section 2.4 describes a two-tier framework in which constituents learn at the lower level and a resource manager is trained with Deep Q-learning to allocate resources. Section 3.3 likewise places resource-manager training at Tier II, supporting the claimed similarity to a two-layer scheme. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“While information disclosure has been studied as a static design choice [28,29], and network structure has been recognized as influencing information flow, the use of dynamically adaptive, RL-driven information modulation as a standalone governance lever represents a novel paradigm.”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > Why Information Modulation as a Governance Mechanism?
What the sources say
- Could not verifycites [28] Stewart, 2016, Small Groups and Long Memories Promote Cooperation, Sci. Rep. sourceExperimental — accuracy not yet measured
The citation is attached to the statement that information disclosure has been studied as a static design choice, not to the citing authors’ claim of novelty. The Methods section examines memory-dependent strategies in public-goods games, and the Discussion addresses group size and cooperation; neither addresses static information-disclosure design. Searches for information disclosure and adaptive information modulation found no supporting passage in our copy (page numbers unavailable). (LLM, conf=0.76)
- Supportedcites [29] Bradley, 2018, Does Observability Affect Prosociality?, Proc. R. Soc. B: Biological Sci. source
The citation is attached to the background claim that information disclosure has been studied as a design choice, not to the citing authors’ claim that RL-driven modulation is novel. The Introduction (Theoretical moderators; page unavailable) describes experimental manipulations of observability, and the Discussion (Theoretical implications; page unavailable) describes studies making participants’ contribution histories observable to others. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
79 more claims
Create a free account to see all 145 claims and every source verdict.
- Probably not a substantial claim. 31
- It may state two separate claims as one. 36
- The quote only partly states this claim. 2
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 3.4. RL Manager
“The cooperation rate for agent i when interacting with agent j is ar ij = 1 t t k=1 I(a k ij = C).”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 7. Prompts Utilized in This Work
“{your ratio} will indicate the agent's own overall cooperation rate with all other agents from previous interactions;”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 9.2. Influence of Quantitative Cooperation Rate on LLM's Behavior
“where the cooperation ratio of opponent is fixed to 75% in Table 3 and the cooperation ratio of LLM agents themselves is fixed to 75% in Table 4.”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > Additional information
“Supp.Table 3: The percentage of LLM agents choosing C when they get access to different cooperation ratios of themselves and their coplayers.The cooperation ratios of LLM agents themselves are fixed at 75% in this table.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > Supplemental Material 6. Overview on Game Theory and Social Dilemmas
“Games belonging to this category all have the following matricial form: Supp.Table 1: Canonical payoff matrix for a symmetric 2-player social dilemma game.C D C (R,R) (S, T) D (T, S) (P, P)”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > Supplemental Material 6. Overview on Game Theory and Social Dilemmas
“The first element of the vector corresponds to the utility gained by the first (or column) player, while the second element reflects the utility earned by the second (or row) player.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 3.2. POMDP and Actor-Critic
“The actor is updated by performing gradient ascent on the expected return J(θ), using the policy gradient theorem:”
§ Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent AI Systems > 8. Background of POMDP and Actor-critic
“The actor updates its policy according to gradient descent, ∇ θ log π(a|s; θ)A(s, a), where A(s, a) is the advantage function, calculated by A(s, a) = r + γV (s ′ ) -V (s).”
Citations
Cites
Reinforcement Learning: An Introduction
Sutton, R. S., & Barto, A. G. (1998) · IEEE Transactions on Neural Networks · doi:10.1109/tnn.1998.712192 · Cited by 27,225
Statistical Power Analysis for the Behavioral Sciences
Cohen, J. (2013) · doi:10.4324/9780203771587 · Cited by 25,373
Governing the Commons: The Evolution of Institutions for Collective Action
Ostrom, E. (1990) · doi:10.1017/cbo9780511807763 · Cited by 20,653
Axelrod, R., & Hamilton, W. D. (1981) · Science · doi:10.1126/science.7466396 · Cited by 20,065
Five Rules for the Evolution of Cooperation
Nowak, M. A. (2006) · Science · doi:10.1126/science.1133755 · Cited by 5,894
Chain-Of-Thought Prompting Elicits Reasoning in Large Language Models
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q. V., & Zhou, D. (2022) · Advances in Neural Information Processing Systems 35 · doi:10.52202/068431-1800 · Cited by 4,505
Connolly, T., & Axelrod, R. (1984) · Administrative Science Quarterly · doi:10.2307/2392957 · Cited by 3,890
Kivela, M., Arenas, A., Barthelemy, M., Gleeson, J. P., Moreno, Y., & Porter, M. A. (2014) · Journal of Complex Networks · doi:10.1093/comnet/cnu016 · Cited by 3,235
The structure and dynamics of multilayer networks
Boccaletti, S., Bianconi, G., Criado, R., del Genio, C. I., Gómez-Gardeñes, J., Romance, M., Sendiña-Nadal, I., Wang, Z., & Zanin, M. (2014) · Physics Reports · doi:10.1016/j.physrep.2014.07.001 · Cited by 3,125
MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework
Hong, S., Zhuge, M., Chen, J., Zheng, X., Cheng, Y., Zhang, C., Wang, J., Wang, Z., Yau, S. K. S., Lin, Z., et al. (2023) · arXiv · doi:10.48550/arxiv.2308.00352 · Cited by 2,825
Research works printed but not linked: 2
- Sutton, 2018, Reinforcement Learning: An Introduction
- Konda, 1999, Actor-Critic Algorithms, Adv. Neural Inform. Process. Syst., 12, 1008
Cited by
Chen, X., & Cheng, K. (2026) · Machines · doi:10.3390/machines14070787
Çetinkaya, İ. O., Khodadadian, S., & Topcu, T. G. (2026) · Procedia Computer Science · doi:10.1016/j.procs.2026.08.188 · Cited by 1
Digital Twins and Agentic AI: Applications, Risks, and Governance Frameworks
Alnoman, A. A., Khwaja, A. S., Anpalagan, A., & Woungang, I. (2026) · Future Internet · doi:10.3390/fi18100501
Metadata
- Authors
- Chen, Q., Ilami, S., Lore, N., & Heydari, B.
- Affiliations
- Northeastern University (US)
- Northeastern University Department of Mechanical and Industrial Engineering, Institute of Experiential AI, and Network Science Institute, , ,
- ORCID
- Heydari, Babak
- Venue
- Journal of Mechanical Design
- Publisher
- ASME International
- Published
- 2026-03-09
- Volume
- 148
- Issue
- 4
- ISSN
- 0161-8458 (ISSN-L)
- 1050-0472
- 1528-9001
- OpenAlex
- W7118190101
- Cited by
- 3
- References
- 52
- Citations by year
- 2026: 3
- FWCI
- 10.39
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- Journal of Mechanical Design (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Multi-Agent Systems and Negotiation
- Social Sciences › Decision Sciences › Management Science and Operations Research › Game Theory and Applications
- Social Sciences › Sociology and Political Science › Evolutionary Game Theory and Cooperation
- Keywords
- adaptive governance, multi-agent systems, agentic AI, large language models, information transparency, information disclosure, reinforcement learning, cooperation, prisoner's dilemma, socio-technical systems, agent autonomy, information flow networks, human-robot collaboration
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-01 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent Artificial Intelligence Systems. https://app.cochat.ai/papers/10.1115/1.4070755
- The paper
- Chen, Q., Ilami, S., Lore, N., & Heydari, B. (2026). Adaptive Information Modulation: Designing Governance Mechanisms for Multi-Agent Artificial Intelligence Systems. Journal of Mechanical Design. https://doi.org/10.1115/1.4070755
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.