Source-linked AI summary
Optimal Networks with Accumulative Costs and Bidirectional Communication
Juan M. C. Larrosa, Fernando Tohmé
TL;DR
The paper asks how equilibrium topology changes when information flows bidirectionally but link costs remain initiator-funded. It extends a strategic network-formation model and finds that strict Nash networks are sequential lines with alternating active nodes, a structure that interrupts cost accumulation and is efficient and Pareto optimal.
Problem
The paper examines which network structure can emerge as a stable and optimal equilibrium when information flows bidirectionally but link costs remain asymmetric.
Method
The paper extends Larrosa and Tohmé’s strategic network-formation approach with directed, initiator-funded links that provide symmetric information benefits.
Results
Strict Nash equilibria are sequential line networks with even or odd active nodes, and these networks are efficient and Pareto optimal.
Takeaways & Limitations
Intermediate active agents connecting to their two immediate neighbors interrupt cost accumulation and support the welfare-maximizing equilibrium structure.
Abstract
from arXiv · showhide
This paper is an extension of the approach of Larrosa and Tohmé (2003), in which the payoff function is modified by allowing information to flow in both directions. Costs continue to be paid by the agent who initiates the connection, so this asymmetry reveals changes in the final equilibrium topology. We find that several optimal topologies persist as Nash networks, but that a strict Nash network corresponds to the sequential line network with intermediate active nodes; that is, agents are placed in a line and, every other agent, they connect with the predecessor agent and with the subsequent agent. In this way, cost accumulation is interrupted and benefits are maximized.
Introduction
The paper applies a strategic, noncooperative network-formation framework to bidirectional information flow with initiator-funded links. It asks which topology maximizes information access with minimal connections and how such structures relate to equilibrium and optimality.
- Introduction: The paper uses a noncooperative framework in which agents choose connection strategies and resulting rational decisions generate Nash equilibria.The approach focuses on how networks form and which structures are stable or efficient.
- Introduction: The model extends Larrosa and Tohmé (2003) by allowing information to flow bidirectionally while link costs remain paid by the initiating agent.The paper therefore studies symmetric information benefits alongside asymmetric connection costs.
- Introduction: The central question is which optimal communication topology supports bidirectional flow under accumulative costs.The paper seeks a structure that combines few connections with broad information access.
- Introduction: The paper is organized around bidirectional-flow definitions, the model, equilibrium architecture, stability and optimality criteria, and comparison with the unidirectional case.The final section discusses analogies and differences with the earlier setting.
I. The Case of Bidirectional Information Flow
The paper models networks as directed graphs whose links transmit information in both directions, while the initiating agent alone pays the connection cost. This benefit-cost asymmetry changes optimal structures relative to unidirectional flow.
- I. The Case of Bidirectional Information Flow: Bidirectional information flow gives both connected agents access to the other’s information, but the initiator alone finances the link.The graph remains directed because it records who establishes and pays for each connection.
I.1 Definitions
The definitions specify agents’ binary link-building strategies, paths and their lengths, information transmission assumptions, and the exclusion of empty-network equilibria. Information is assumed to transmit without distortion or decay.
- I.1 Definitions: A direct link initiated by i gives i access to j’s information and also gives j access to i’s information and i’s contacts.The convention assumes each agent’s information is sufficiently valuable to rule out the empty network as an equilibrium.
- I.1 Definitions: Each agent chooses a binary strategy vector whose entries indicate whether she establishes direct links with other agents.The analysis is restricted to pure strategies, and self-links are excluded.
- I.1 Definitions: A path is a sequence of distinct agents linked according to the joint strategy, and its length equals the number of intermediate links.A direct link is a path of length 1.
- I.1 Definitions: Information transmission is assumed to be distortionless and nondecaying, so an agent’s information value is unchanged whether access is direct or indirect.The value of information also does not change as more agents learn it or as an individual learns more information.
- I.1 Definitions: The bidirectional-flow case is defined so that information travels both ways once contact is established.This provides the communication setting used by the subsequent model.
II. The Model
The model represents bidirectional communication with directed links that encode who pays for initiation, while closure and path-access definitions determine which agents obtain information. Examples contrast bidirectional and unidirectional access.
- II. The Model: Bidirectional-flow networks give connected agents symmetric information benefits but preserve directed cost responsibility for the link initiator.Thus gij = 1 and gji = 1 are not equivalent: they identify different financiers even though communication benefits are shared.
- II. The Model: Strict Nash equilibria with the minimum number of links support a peripherally financed star network that is stable and optimal.This result is stated as the equilibrium outcome for the model’s payoff function.
- II. The Model: A strategy profile is represented as a directed graph in which nodes are agents and links encode initiation, while information circulates in both directions.The associated tables and figures represent the profile and its network graph.
- II. The Model: The model counts links in paths that terminate at an agent and allows more than one path between two agents.These path-based quantities support the derivation of agents’ information access and costs.
- II. The Model: Under bidirectional information, every agent in the Example 2 network accesses every agent’s information, unlike the unidirectional case where access is progressively narrower.The bidirectional configuration yields universal access, whereas agents 4 and 5 in the unidirectional example access only their own information.
II.1 Connections and Benefits
The model distinguishes direct and indirect connections while allowing information to flow bidirectionally. Agents access network information through connections, but the initiating agent finances the accumulated costs of the paths used.
- Agent i’s accessible-information set combines directly linked agents with agents reached through paths of value greater than one.The first term in N_i;g captures direct links, while the second captures indirect links.
- Accumulated costs include every direct and indirect link in the paths toward accessed information, so longer paths become more expensive as intermediaries are added.Each link has unit cost, and intermediaries must be paid as information passes through them.
- Bidirectional communication lets connected agents access each other’s information, while only the link initiator pays the connection cost.The graph is directed for cost accounting even though benefits are symmetric between connected agents.
- Under bidirectional flow, an agent can access all network members’ information when connected through the network, regardless of which agent initiated each link.This differs from the unidirectional case, where accessed information depends on flow direction.
- The payoff compares strategies by balancing the amount of information obtained against the number of financed links required to reach it.With equal information, fewer required links are preferred; with equal link requirements, greater information is preferred.
III. Equilibrium and Optimality
The section characterizes equilibrium networks under bidirectional information flow with initiator-paid, accumulative costs. Sequential-connection line networks with alternating active nodes are exactly the strict Nash networks with the minimum number of links, while other optimal structures can remain Nash networks.
- Equilibrium characterization: In these line networks, odd or even nodes initiate links to their immediate neighbors while the remaining nodes remain passive.The two activation patterns differ only in which parity of interior nodes is active, with endpoint adjustments depending on network size.
- Equilibrium characterization: The proof uses unilateral deviations: disconnecting saves connection costs but sacrifices network-wide information, while nonsequential links either disconnect the network or leave intermediate agents isolated.Passive agents cannot improve their benefit because they already access all information without paying costs.
- Optimality: Every sequential line network and the centrally supported star network attain the maximum net social benefit under bidirectional information.The section favors the line configuration because it has a symmetry criterion that the centrally supported star lacks.
- Alternative Nash networks: A centrally supported star can be Nash and Pareto efficient, but one active central agent bears the network’s costs while passive agents receive all information for free.This cost asymmetry can make the central agent prefer another structure providing the same information at lower cost.
- Equilibrium characterization: Sequential-connection line networks with either even or odd active nodes are strict Nash networks with the minimum number of links.This is stated as an if-and-only-if characterization.
III.4 Social Welfare
Sequential line networks with even or odd active nodes are both efficient and Pareto optimal. They share maximum social benefit with the centrally supported star, while retaining a symmetry criterion that favors the linear configurations.
- The sequential line network is stable because agents have no incentive to cut links or form new ones once it emerges.
- Efficiency compares total social welfare across networks, whereas Pareto optimality requires that no alternative improve one agent without harming another.
- A sequential line network with even or odd active nodes that is strict Nash is both efficient and Pareto optimal.
- All sequential line networks share the maximum benefit with the centrally supported star network.
Conclusion
Accumulative costs constrain broad strict Nash networks, favoring sequential linear networks with alternating active nodes. A proposed extension could qualify links and potentially support broader strict Nash topologies.
- Accumulated costs restrict broad strict Nash networks, while alternating active nodes interrupt cost accumulation and maximize social welfare.
- Qualifying link creation more than passive acceptance is proposed as a future extension that could support broader strict Nash topologies.