Dynamic Network Formation and Coalition Building through Real-Time Communication

Last registered on September 29, 2025

View Trial History

Pre-Trial

Trial Information

General Information

Title

Dynamic Network Formation and Coalition Building through Real-Time Communication

RCT ID

AEARCTR-0016865

Initial registration date

September 24, 2025

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

September 29, 2025, 10:54 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Country

China

Region

Guangdong

Primary Investigator

Name

Deng Xiaotian

Affiliation

The Hong Kong University of Science and Technology (Guangzhou)

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Chenyang Li

PI Affiliation

Hong Kong University of Science and Technology(GuangZhou)

Contact Investigator

Additional Trial Information

Status

In development

Start date

2025-10-30

End date

2026-01-30

Keywords

Behavior, Lab, Other

Additional Keywords

experimental economics, network games, coalition formation, real-time communication, strategic behavior, endogenous networks, social networks, game theory, multi-agent systems, behavioral economics, laboratory experiment, network formation, strategic communication, alliance formation, economic incentives

JEL code(s)

C92, D85, C71, D91, D23

Secondary IDs

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

This study investigates how individuals form coalitions and make strategic decisions in network-based economic environments through a novel experimental design that combines real-time communication with endogenous network formation. Participants engage in a multi-round experimental game where they can communicate freely through instant messaging and strategically self-select their network connections while competing in performance-based tasks.

The experiment examines how different network structures, economic incentives, and communication opportunities influence coalition formation, strategic communication patterns, and individual decision-making processes across multiple rounds of interaction. By allowing participants to endogenously form their social networks through friend requests and maintain real-time conversations, this research provides unique insights into the organic development of strategic alliances in competitive economic environments.

The study addresses fundamental questions about social network effects on economic behavior, including how network topology affects cooperation and competition, how real-time communication influences trust and coordination, and how individuals navigate the tension between individual performance and collective action. The experimental design offers methodological innovations by integrating dynamic network formation with strategic communication in a controlled laboratory setting.

The findings will contribute to our understanding of social network effects on economic behavior, strategic interaction in multi-agent systems, and the role of communication in facilitating or hindering cooperative behavior. Results may have broader implications for organizational design, team formation, and policy interventions in networked environments where individuals must balance competitive and cooperative incentives.

External Link(s)

Registration Citation

Citation

Li, Chenyang and Deng Xiaotian. 2025. "Dynamic Network Formation and Coalition Building through Real-Time Communication." AEA RCT Registry. September 29. https://doi.org/10.1257/rct.16865-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Participants engage in a multi-round network-based economic game involving task performance and strategic communication. The intervention manipulates network structure (allowing participants to form connections endogenously) and communication opportunities (real-time messaging capabilities) while participants compete in performance-based activities. Different experimental conditions vary the network formation rules and incentive structures to examine how these factors influence coalition formation and strategic behavior. Each session consists of multiple rounds where participants make individual performance decisions while having opportunities for communication and network formation with other participants.

Intervention (Hidden)

The experiment implements a 5-round network game where 7 participants simultaneously complete cognitive tasks (matrix calculations) while engaging in strategic communication and network formation. Participants can send friend requests and communicate via real-time chat to form coalitions.

Key experimental manipulations include:
1. Network Structure: Three conditions (star, ring, full connectivity) determining initial network topology and connection constraints
2. Coalition Threshold: Varies from 2-7 participants required for successful coalition formation
3. Economic Incentives: Performance-based earnings from individual tasks combined with potential coalition bonuses and reporting rewards
4. Communication: Unrestricted real-time messaging between connected participants
5. Network Formation: Participants can endogenously modify network structure through friend requests

Each round lasts 10 minutes with participants earning points based on task performance, coalition membership outcomes, and strategic reporting decisions. The design examines how network constraints and communication opportunities affect the emergence of cooperative and competitive behaviors, with particular focus on coalition formation dynamics and information sharing patterns. Treatment conditions systematically vary network structure and threshold parameters to identify causal effects on strategic behavior and alliance formation.

Intervention Start Date

2025-10-30

Intervention End Date

2025-12-31

Primary Outcomes

Primary Outcomes (end points)

1. Coalition Formation Rate: Proportion of successful coalitions formed across different network structures and threshold conditions

2. Network Connection Choices: Pattern of endogenous network formation including frequency and targeting of friend requests

3. Strategic Communication Behavior: Volume, timing, and content patterns of real-time messaging between participants

4. Reporting Decisions: Individual choices to report or participate in coalition activities under different incentive structures

5. Individual Task Performance: Accuracy and completion rates in cognitive tasks across different social and network contexts

6. Earnings Distribution: Total points/payoffs earned by participants across individual performance, coalition participation, and strategic decision components

Primary Outcomes (explanation)

Coalition Formation Rate will be constructed as the percentage of rounds in which participants successfully form coalitions meeting the threshold requirements, measured separately by network structure type.

Strategic Communication Behavior will be operationalized through multiple constructed variables including: (a) total message volume per participant per round, (b) communication network density (proportion of possible connections used for messaging), and (c) temporal communication patterns (timing of messages relative to task deadlines and coalition formation opportunities).

Network Connection Choices will be measured as endogenous network formation patterns, constructed from the sequence and success rate of friend requests, final network configurations, and network stability across rounds.

Reporting Decisions will be coded as binary choices (report/participate) in coalition opportunities, with additional analysis of decision timing and relationship to communication patterns.

Earnings Distribution will combine multiple payment components weighted according to experimental parameters: individual task performance scores, coalition membership bonuses (when applicable), and reporting rewards, providing a comprehensive measure of strategic success across different behavioral dimensions.

Secondary Outcomes

Secondary Outcomes (end points)

1. Learning Effects: Changes in task performance and strategic behavior across experimental rounds

2. Trust and Reciprocity Measures: Patterns of mutual friend acceptance, message reciprocity, and cooperative behavior sustainability

3. Network Centrality Effects: Relationship between participants' network positions and their behavioral outcomes

4. Communication Content Analysis: Qualitative categorization of message types (coordination, deception, information sharing, negotiation)

5. Decision Timing Patterns: Response latencies in task completion, coalition decisions, and communication initiation

6. Risk Preferences Revelation: Behavioral indicators of risk attitudes through coalition participation decisions under uncertainty

7. Social Preference Indicators: Evidence of altruistic, competitive, or reciprocal motivations in strategic choices

8. Adaptation Strategies: Changes in network formation and communication strategies in response to previous round outcomes

Secondary Outcomes (explanation)

Learning Effects will be measured as within-participant changes in performance metrics across rounds, controlling for network structure and threshold conditions. This includes both task-specific learning (improvement in cognitive task accuracy) and strategic learning (optimization of coalition and communication strategies).

Trust and Reciprocity Measures will be constructed from behavioral indicators including: mutual friend request acceptance rates, reciprocal messaging patterns, and sustained cooperative relationships across multiple rounds. These will be operationalized as network-level trust indices and individual reciprocity scores.

Network Centrality Effects will analyze the relationship between participants' structural positions (degree centrality, betweenness centrality, closeness centrality) and their outcomes, examining whether central network positions confer strategic advantages.

Communication Content Analysis will employ text analysis methods to categorize messages into functional types: coordination attempts, strategic deception, information sharing, negotiation, and social chat. This will be coded both manually and through automated text classification.

Decision Timing Patterns will measure response latencies and temporal coordination, including time-to-decision in coalition choices, communication response times, and task completion speeds as indicators of cognitive load and strategic deliberation.

Social Preference Indicators will be inferred from choices that deviate from pure profit maximization, including coalition choices that benefit others, information sharing that reduces competitive advantage, and reporting decisions that reflect social motivations beyond individual payoffs.

Experimental Design

This study employs a randomized controlled laboratory experiment using a between-subjects design with multiple treatment conditions. Participants are randomly assigned to groups and engage in a multi-round network-based economic game. The experimental design systematically varies network structure constraints and economic incentive parameters to examine their effects on coalition formation and strategic communication.

The experiment uses a factorial design manipulating: (1) network formation rules that determine how participants can connect with each other, and (2) threshold parameters that affect coalition success requirements. Each experimental session involves multiple rounds of interaction where participants complete individual performance tasks while having opportunities for strategic communication and network formation.

Randomization occurs at both the session level (treatment assignment) and within-session level (certain game parameters). The design controls for potential confounds through standardized instructions, identical task structures across conditions, and balanced treatment allocation. Sessions are conducted in a controlled laboratory environment with computer-mediated interaction to ensure treatment fidelity and data quality.

Experimental Design Details

The experiment implements a 2×5 factorial design with 7 participants per session across 5 rounds. The main factors are:

Factor 1 - Network Structure (3 levels):
- Star network: One central node connected to all others; peripheral nodes cannot connect directly
- Ring network: Each participant connected to exactly two neighbors in circular arrangement
- Full network: All participants can potentially connect to all others

Factor 2 - Coalition Threshold (5 levels): 2, 3, 4, 5, 6, or 7 participants required for successful coalition formation

Additional experimental variations include star network bonus conditions (with/without dividend distribution to center node).

Procedure: Each 10-minute round consists of: (1) Network formation phase where participants send/accept friend requests, (2) Task performance phase involving matrix calculation problems, (3) Coalition formation and communication through real-time messaging, (4) Reporting decisions for non-coalition members.

Randomization: Treatment conditions are randomly assigned at session level. Within sessions, task order and certain coalition opportunities are randomized. Initial network structures (where applicable) are randomly assigned while maintaining treatment integrity.

Sample size: Planned 30-40 sessions (210-280 participants total) with balanced allocation across treatment cells. Power analysis indicates adequate sample for detecting medium effect sizes in primary outcomes.

Control measures: Standardized computerized interface, identical cognitive tasks across conditions, controlled laboratory environment, balanced demographic recruitment, and systematic debriefing protocols.

Randomization Method

Computer-based randomization using pseudo-random number generator in controlled laboratory setting

Randomization Unit

Two-level randomization:
1. Session-level randomization: Experimental sessions (groups of 7 participants) are randomly assigned to treatment conditions (network structure × coalition threshold combinations)
2. Individual-level randomization: Within sessions, participants are randomly assigned to initial network positions (where applicable) and certain game parameters such as task presentation order and coalition opportunity sequences are randomized

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

86-100 experimental sessions

Sample size: planned number of observations

600-700 participants

Sample size (or number of clusters) by treatment arms

Network Structure Treatment Arms:
- Star Network: 29-34 sessions (203-238 participants)
- Ring Network: 29-33 sessions (203-231 participants)
- Full Network: 28-33 sessions (196-231 participants)

Within each network structure, sessions are further allocated across 5 coalition threshold conditions (2, 3, 4, 5, 6, 7 participants) with approximately 6-7 sessions per threshold level.

Additional star network bonus conditions: 10-15 sessions allocated to star network with center node dividend distribution.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Primary outcome: Coalition Formation Rate - Unit: Proportion of successful coalitions per session - Expected baseline standard deviation: 0.25 (based on pilot data) - Minimum detectable effect size: 0.15 percentage points (0.6 standard deviations) - Statistical power: 80% at α = 0.05 - Accounts for clustering with intracluster correlation (ICC) = 0.05 Secondary outcomes: Communication behavior measures - Unit: Messages per participant per round - Expected standard deviation: 3.2 messages - Minimum detectable effect size: 1.6 messages (0.5 standard deviations) - Clustering adjustment applied with design effect = 1.2

Supporting Documents and Materials

IRB