Publication Date
Spring 5-6-2022
Document Type
Conference Poster
First Advisor
Murphy, John
Degree Name
B.S. (Bachelor of Science)
Department
Department of Computer Science
Abstract
Loot boxes are digital treasure chests that players spend real money to purchase, wherein the contents are randomly generated. Since players spend money on the pretense they might receive something valuable, many comparisons have been drawn to gambling behavior as the reward is up to chance. To explore this phenomenon, agent-based modeling will be used to simulate this behavior. Agent-based modeling allows us to create heterogenous agents who follow simple rules so that we may observe emergent behavior in a population. An agent-based model was created using Repast Simphony for this end.
Parameters included the player’s internal decision strategy around purchases, the number of players, the network type used to generate connections between player nodes, and the manipulation strategy used by the game. Batch runs of the model were conducted to sweep over these parameters and collect data regarding the purchase trends. Lattice networks lead to more isolated players, where game manipulations were predictably effective on keeping the players habitually spending. Watts networks lead to less isolation, but interestingly the manipulations reduced player likelihood to spend large amounts of money on a new loot box. Random density networks lead to the most interconnected nodes, where interactions between nodes lead to the shortest purchase frequency time at less than half a tick of runtime.
Network structure influenced player behavior more than decision strategy or manipulation used. Players who were totally isolated made fewer purchases than those who were well connected, and were not impacted by the manipulations as sharply as their peers. Though further research and optimizations to the model are needed, it does appear from the model that a player’s social network has strong influence over their likelihood to engage in and fall victim to gambling-like behavior with in-game purchases.
Recommended Citation
Zayed, Lila, "JLootBox: An Agent-Based Model of Social Influence and Gambling in Online Video Games" (2022). Honors Capstones. 1420.
https://huskiecommons.lib.niu.edu/studentengagement-honorscapstones/1420
Comments
2022 CURE Award Winner - 1st Place in Mathematics and Computations Sciences Posters
If you'd like to download and experiment with the model, please follow the installation instructions detailed in the full text for this research project.