Crowdsearch
summary
The gist
A common economic process is crowdsearch, wherein a group of agents is invited to search for a valuable object…
In short
The paper models crowdsearch, where agents decide whether to search for a valuable item based on their search costs. It shows that increasing the number of agents doesn't guarantee success because competition can reduce individual incentives. The probability of finding the object can eventually increase or decrease depending on how agent participation responds to changes in the cost distribution.
Key concepts
- Crowdsearch Model
- This is a game-theoretic model where many agents compete to find a valuable item, like a software bug. Each agent faces a decision: search or not search, balancing the potential reward against the cost of searching and competition from others.
- Equilibrium Threshold c*
- This is the specific cost level that determines whether an agent chooses to participate in the search. It is found by solving a mathematical equation derived from balancing participation rates and costs. This threshold changes based on how many agents are present and the value of the prize.
- Crowding-out Effect
- This effect occurs when inviting more agents lowers individual incentives to search, even if more people are available. This happens because increased competition for the prize reduces the potential payoff for each individual agent, potentially decreasing overall success probability.
- Lower-tail Behavior of cf(c)
- This refers to how the distribution of search costs behaves when costs are very low. The specific pattern in this 'tail' dictates whether the probability of success eventually increases or decreases as more agents are added to the search group.
Terminology used across episodes
This episode discusses
The paper
Hans Gersbach, Akaki Mamageishvili, Fikri Pitsuwan
KOF Swiss Economic Institute · Offchain Labs · Jeb E. Brooks School of Public Policy
A common economic process is crowdsearch, wherein a group of agents is invited to search for a valuable object, such as a software vulnerability or a proof-of-work nonce. We study a binary model of crowdsearch in which agents with heterogeneous search costs decide whether or not to search. Inviting more agents directly expands the pool of potential searchers, but it also increases competition for the prize and lowers individual search incentives. This crowding-out effect may dominate, so that inviting more agents can lower the probability of finding the object. We show that inviting an unlimited crowd need not guarantee success. We also provide sufficient conditions, based on the lower-tail behavior of the cost distribution, under which the probability of success is eventually increasing or eventually decreasing in the number of agents.
Transcript
Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.
Ruth: I'm Ruth.
Warren: And I'm Warren, here with Iris.
Ruth: We're looking at the paper: "Crowdsearch".
Warren: Here's the gist of the paper. A common economic process is crowdsearch, wherein a group of agents is invited to search for a valuable object, such as a software vulnerability or a proof-of-work nonce This paper provides a simple game-theoretic model of crowdsearch, abstracting from the details of particular applications It shows that inviting an unlimited crowd need not guarantee success.
Ruth: Model Setup and Equilibrium Conditions The paper models crowdsearch as a simple contest where each agent makes a binary costly decision whether or not to search The payoff of agent i is defined as ui(si, s−i, ci) = si (p(s−i)V − ci).
Warren: The main analysis characterizes the Bayesian Nash equilibrium in a symmetric threshold strategy where all agents use the same strategy of the form σ(ci) = (1 if ci ≤ cˆ 0 if ci > cˆ for some threshold ˆc).
Ruth: The unique symmetric equilibrium threshold c∗ ≡ c∗(V, q, n) is the solution to cˆ = V Φ(ˆc, q, n), where Φ(ˆc, q, n) ≡ P(ˆc, q, n) / (nF(ˆc)).
Paper summary: Ruth: To summarize what this paper "Crowdsearch" does, the authors provide a simple game-theoretic model of crowdsearch, abstracting away the specifics of any particular application.
Warren: It shows that inviting an unlimited crowd doesn't automatically mean success is guaranteed. They also provide some conditions based on how the costs are distributed that tell us when the chance of finding it will eventually increase or eventually decrease as you add more agents.
Iris: The model focuses on how the size of the invited crowd affects this probability of success, specifically looking at what happens to the equilibrium threshold and that success probability as you get a larger crowd.
Ruth: They identify two main forces that determine which effect dominates—the level of participation in the search versus how responsive participation is to changes in the cost threshold.
Warren: The authors show that if you look at very large contests, the equilibrium threshold settles down to some value, and if it’s positive, the expected number of agents who actually search stabilizes at a constant.
Iris: That convergence point leads to a limit for success probability based on that stable participation level and the cost structure.
Conclusion: Ruth: So looking at "Crowdsearch," what it really boils down to is this fundamental tradeoff between inviting more potential searchers and making sure those individuals still have a strong incentive to actually participate in the search.
Warren: The paper shows that this tradeoff is real, and there are specific conditions based on the cost distribution—specifically how those costs behave at the lower tail—that predict whether adding more people will eventually help or hurt your chances of success.
Iris: The authors conclude by providing sufficient conditions for when the probability of success will eventually increase or decrease in the number of agents, which gives a more nuanced view than just saying "more is better."
Ruth: It really highlights that you can't just keep inviting people hoping they'll help if the structure of your search contest causes them to stop searching altogether.
Warren: The implications are that for any real-world crowdsearch contest, like bug bounties or proof-of-work verification, you need to be aware of how your agent costs interact with the overall pool size.
More episodes
- Daily Summary for 2026-11-19
- 2608.21274-Recommendation Quality and the Concentration of Consumption: Experimental Evidence from Netflix
- 2408.02757-A nonparametric test for diurnal variation in spot correlation processes
- 2306.16393-High-Dimensional Canonical Correlation Analysis
- 2605.17299-Geometric Brownian motion with intermittent entries and exits
- 2607.04753-Fooling Yourself: how narratives shape beliefs
- 2408.14671-Inference on Treatment Effects with Repeatedly Mismeasured High-Dimensional Controls
- 2201.06898-Difference-in-Differences Estimators for Treatments Continuously Distributed at Every Period
- 2501.15307-Influence Functions: A Unified Differentiation-Based Framework
- The day in conversation for 2026-11-19