Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Modeling Human Behavior in a Strategic Network Game with Complex Group Dynamics

Jonathan Skaggs, Jacob W. Crandall

TL;DR

This work tackles modeling human decision-making in strategic networks using the Junior High Game (JHG). It systematically compares behavior-matching TFT and community-aware CAB models, learning parameterizations via PSO and EPDM to capture mean vs distribution of behavior. The standout finding is that CAB with EPDM (hCAB-EPDM) best reproduces human population dynamics in small groups, and a user study shows individual hCAB agents are difficult to distinguish from humans, suggesting plausible individual behavior. The results demonstrate a data-efficient, interpretable route to simulating human networks and provide a baseline for future neural or hybrid approaches in strategic interactions.

Abstract

Human networks greatly impact important societal outcomes, including wealth and health inequality, poverty, and bullying. As such, understanding human networks is critical to learning how to promote favorable societal outcomes. As a step toward better understanding human networks, we compare and contrast several methods for learning, from a small data set, models of human behavior in a strategic network game called the Junior High Game (JHG). These modeling methods differ with respect to the assumptions they use to parameterize human behavior (behavior vs. community-aware behavior) and the moments they model (mean vs. distribution). Results show that the highest-performing method, called hCAB, models the distribution of human behavior rather than the mean and assumes humans use community-aware behavior rather than behavior matching. When applied to small societies (6-11 individuals), the hCAB model closely mirrors the population dynamics of human groups (with notable differences). Additionally, in a user study, human participants were unable to distinguish individual hCAB agents from other humans, thus illustrating that the hCAB model also produces plausible (individual) human behavior in this strategic network game.

Modeling Human Behavior in a Strategic Network Game with Complex Group Dynamics

TL;DR

This work tackles modeling human decision-making in strategic networks using the Junior High Game (JHG). It systematically compares behavior-matching TFT and community-aware CAB models, learning parameterizations via PSO and EPDM to capture mean vs distribution of behavior. The standout finding is that CAB with EPDM (hCAB-EPDM) best reproduces human population dynamics in small groups, and a user study shows individual hCAB agents are difficult to distinguish from humans, suggesting plausible individual behavior. The results demonstrate a data-efficient, interpretable route to simulating human networks and provide a baseline for future neural or hybrid approaches in strategic interactions.

Abstract

Human networks greatly impact important societal outcomes, including wealth and health inequality, poverty, and bullying. As such, understanding human networks is critical to learning how to promote favorable societal outcomes. As a step toward better understanding human networks, we compare and contrast several methods for learning, from a small data set, models of human behavior in a strategic network game called the Junior High Game (JHG). These modeling methods differ with respect to the assumptions they use to parameterize human behavior (behavior vs. community-aware behavior) and the moments they model (mean vs. distribution). Results show that the highest-performing method, called hCAB, models the distribution of human behavior rather than the mean and assumes humans use community-aware behavior rather than behavior matching. When applied to small societies (6-11 individuals), the hCAB model closely mirrors the population dynamics of human groups (with notable differences). Additionally, in a user study, human participants were unable to distinguish individual hCAB agents from other humans, thus illustrating that the hCAB model also produces plausible (individual) human behavior in this strategic network game.
Paper Structure (36 sections, 25 equations, 12 figures, 4 tables, 2 algorithms)

This paper contains 36 sections, 25 equations, 12 figures, 4 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. A strategic network game to model human networks
  3. Modeling human behavior in the JHG using human data
  4. Parameter-based models of behavior
  5. Learning parameterizations
  6. Experiment A: Evaluating population dynamics
  7. Experiment design
  8. Data sets
  9. Metrics to evaluate population dynamics
  10. Results
  11. Experiment B: Interacting with human players
  12. User study design
  13. Results
  14. Conclusions and delimitations
  15. Overview of the Supplementary Material
  16. ...and 21 more sections

Figures (12)

  • Figure 1: Example scenario illustrating various JHG dynamics. (a) Popularity of the players over time. (b) Token allocations in rounds 1-4. Red dashed arrows indicate tokens used to attack while black solid arrows indicate tokens given. (c) Token allocations in rounds 5-6.
  • Figure 2: Comparisons of human and agent societies. Dots (a-c), numbers (d), and lines (e-f) denote the mean, error ovals (which do not account for covariance) and ribbons show the standard error.
  • Figure 3: User study results comparing humans and hCABs. (a) Mean popularity over time. Error ribbons show the standard error. (b) Percent of tokens used to give, keep, and take by each player type. (c) Percentage of players correctly identified by human participants verses random chance.
  • Figure 4: Illustrative game showing the popularity of players over time (top) and network structure in select rounds (bottom). In the graphs, arrows indicate token transactions in the current round (green = give; red = take), while nodes more connected historically tend to be closer to each other.
  • Figure :
  • ...and 7 more figures