Perception Games and Privacy

Ronen Gradwohl, Rann Smorodinsky

TL;DR

This paper introduces "perception games" to model how individuals' privacy concerns influence their strategic decisions. It shows that these concerns can lead to pooling behavior (hiding one's true type) and explores the conditions under which this occurs, offering a new framework for understanding privacy in strategic interactions.

What Problem Does This Paper Address?

The paper tackles the question of how privacy concerns affect decision-making in strategic settings. While traditional game theory assumes players act solely based on material payoffs, this paper recognizes that individuals often care about what others infer from their actions. The authors investigate whether privacy concerns lead to "pooling behavior" (where individuals with different types act the same way to conceal their identity) and whether this pooling results in social inefficiency. They aim to provide a formal model that captures this "privacy-aware" decision-making, moving beyond purely material considerations.

Methods

The core contribution is the introduction of perception games, a variant of signaling games. In these games, a decision-maker (DM) chooses an action, which is observed by a "Big Brother" (BB). The DM's utility depends not only on the action and their type but also on BB's belief about the DM's type. This belief is formed based on the observed action. The key novelty is incorporating the DM's perception of BB's belief directly into the DM's utility function.

The authors define two types of privacy concerns:

  • Anonymity-seeking: The DM prefers that BB's ex-post belief about their type is the same as the ex-ante belief (i.e., no information leakage).
  • Identification-averse: The DM wants to avoid full revelation of their type, with maximal leakage being the worst outcome.

They then analyze conditions for pooling equilibria (where all types take the same action) for both types of privacy concerns, using formal game-theoretic definitions and proofs.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Privacy as a Strategic Concern: The paper frames privacy not just as a personal preference but as a strategic consideration that influences actions in games.
  • Perception Games: This new framework allows for modeling situations where an agent's utility is affected by what others believe about them, based on their actions.
  • Pooling Behavior: Privacy concerns can incentivize individuals to act identically, even if their underlying types are different, to prevent others from inferring their true nature. This is a direct consequence of wanting to avoid unwanted inferences.
  • "Tyranny of the Majority": In some scenarios, individuals with privacy concerns might be forced to reveal information if a majority of others do not have such concerns, highlighting how social norms can override individual privacy preferences.
  • Welfare Implications: Privacy legislation, which might hide actions, can sometimes increase social welfare by allowing individuals to act optimally without fear of misinterpretation, even if it seems counterintuitive.

Key Results

  • The paper establishes necessary and sufficient conditions for full pooling equilibria in perception games for both anonymity-seeking and identification-averse agents.
  • They demonstrate that privacy concerns do not always imply pooling; non-pooling equilibria can also exist.
  • Through an example, they show that even when individuals have privacy concerns, a lack of privacy legislation can sometimes lead to higher overall welfare, especially in multi-player settings.
  • The paper proves the existence of perception equilibria under certain continuity assumptions on the utility function.

Limitations and Future Work

  • Static Model: The current model is static and one-shot. The authors suggest future work could extend this to dynamic, repeated games to explore how privacy concerns evolve over time and with repeated interactions.
  • Simplified Utility: The model uses a simplified utility function that doesn't account for perceptions of perceptions (unlike psychological games). Future work could explore richer belief structures.
  • Big Brother Interpretation: While the model is technically agnostic, the interpretation of "Big Brother" as a real player versus a figment of imagination has conceptual implications that could be further explored.
  • Generalization: The paper focuses on relatively simple perception games. Extending the framework to more complex scenarios with multiple receivers or different types of privacy concerns is a natural next step.

Summary updated

Abstract

Players (people, firms, states, etc.) have privacy concerns that may affect their choice of actions in strategic settings. We use a variant of signaling games to model this effect and study its relation to pooling behavior, misrepresentation of information, and inefficiency. We discuss these issues and show that common intuitions may lead to inaccurate conclusions about the implications of privacy concerns.
BibTeX
@misc{gradwohl2014perception,
  author = {Ronen Gradwohl and Rann Smorodinsky},
  title = {{Perception Games and Privacy}},
  year = {2014},
  eprint = {1409.1487},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1409.1487},
}
Figure