Strategic Silence and Information Withholding in Competitive Games

18 concepts9 papers1 proposal

In single-agent decision theory, information is a strict economic good. If a lone robot is navigating a maze, giving it a map or real-time sensor updates can only improve its path; at worst, the robot can simply choose to ignore the extra data. More information never degrades performance.

However, when multiple agents interact competitively, this fundamental assumption breaks down. In a game, information is not just a tool for optimization—it is a signal that alters the entire equilibrium of the system. When you acquire information, your opponents know that your strategy will change, and they adjust their own behavior accordingly. This strategic feedback loop can lead to the information paradox: a scenario where having access to more information, or communicating more openly, yields a worse outcome for the informed party or the system as a whole.

To navigate this paradox, competitive agents must treat information disclosure as a scarce, strategic action, often deploying strategic silence—the deliberate withholding of information—to protect their competitive advantage.

Why More Information Can Backfire

To understand why information can be detrimental, we must look at how it shifts Nash equilibria. In dynamic settings, real-time feedback loops can inadvertently intensify competition. As shown by Basar et al., providing players with continuous state updates can drive up the collective "Price of Information," resulting in higher total costs compared to a regime where players commit to blind, pre-planned strategies.

The core mechanism driving this phenomenon is the rational adaptation of competitors. As Hogeboom-Burr et al. demonstrate, while a single actor can ignore data, a team or player in a game cannot prevent opponents from reacting to the knowledge that the data is available. This distinguishes multi-agent strategic environments from single-player control scenarios.

This divergence is highly sensitive to the nature of the game. Yu‐Chi Ho observed that while better measurement quality universally reduces costs in cooperative settings (such as certain Linear-Quadratic-Gaussian games), it creates a zero-sum tug-of-war in competitive duopolies, where variations in measurement quality cause one player's informational gain to directly penalize the other.

Strategic Silence as an Optimal Policy

Because open communication and information abundance can be weaponized by opponents, agents must learn when to speak and when to remain silent.

In multi-agent reinforcement learning, this behavior emerges naturally when communication channels are gated and agents maximize individual rewards. Tushar Jain developed the IC3Net framework to show that when agents compete, they naturally learn to shut down communication channels to prevent information leakage. Rather than oversharing, agents balance coordination against the risk of exposing their internal states or strategies to exploitation.

[Agent State] ---> (Gating Mechanism) ---> [Optional Communication]
                          |
                  Strategic Silence?
                          |
                          v
             (Prevents Opponent Exploitation)

This tactical use of silence extends beyond simple grid-worlds to advanced language models and social networks:

  • Large Language Models: Shi et al. found that in resource-gathering simulations, competitive Large Language Models (LLMs) do not always resort to explicit lying to deceive opponents; instead, they opt for strategic silence. Traditional game-theoretic benchmarks often force agents to communicate continuously, completely missing this sophisticated form of non-verbal deception that easily evades standard message-level filters.
  • Social Networks: In decentralized social environments, Shepherd et al. demonstrated that reinforcement learning agents can maximize their social utility and preserve network connections simply by selectively withholding their opinions rather than broadcasting them publicly.

The Interplay of Information and Resource Constraints

The value of withholding or revealing information is also deeply tied to physical constraints, such as resource endowments. In resource-allocation conflicts like Colonel Blotto games, information asymmetry interacts heavily with absolute strength.

Mahnoosh Alizadeh demonstrates that perfect information about an arena does not guarantee victory if an opponent holds an overwhelming resource advantage. The utility of information is non-monotonic and contingent on whether you have the physical capacity to act on it.

Conversely, committing resources early can act as an accidental information leak. As Vu et al. show, pre-allocating resources creates a stark trade-off: while it provides a baseline advantage, it reveals critical structural information to opponents, stripping the player of strategic flexibility and allowing the adversary to optimize their counter-strategy.

Controlled Revelation

If withholding information preserves strategic flexibility, can revealing information ever be rational? Paradoxically, yes—if it alters the relative distribution of knowledge to an opponent's disadvantage.

Orlandi et al. explored how highly informed agents can be incentivized to participate in joint computations without monetary rewards. They found that an agent may voluntarily reveal private information if doing so diminishes the exclusivity or relative information benefit that a specific competitor stands to gain. By carefully managing how knowledge is distributed across the entire group, agents protect their competitive standing, proving that the value of information is always relative, never absolute.

Key Foundational Work

Prices of Anarchy, Information, and Cooperation in Differential Games (2011) — contributed 2 concept elements to this direction.
Informational properties of the Nash solutions of two stochastic nonzero-sum games (1974) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_strategic_information_withholding_nuance_e07e0e9f,
  title = {strategic information withholding (nuanced value of information)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/strategic-information-withholding-nuanced-value-of-information/rdr_e07e0e9fd6d44a86be4e2a746f8aa891},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_strategic_information_withholding_nuance_e07e0e9f_v1,
  title = {strategic information withholding (nuanced value of information)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/strategic-information-withholding-nuanced-value-of-information/rdr_e07e0e9fd6d44a86be4e2a746f8aa891/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake