DOI: https://doi.org/10.15368/theses.2020.72
Available at: https://digitalcommons.calpoly.edu/theses/2191
Date of Award
6-2020
Degree Name
MS in Computer Science
Department/Program
Computer Science
College
College of Engineering
Advisor
Foaad Khosmood
Advisor Department
Computer Science
Advisor College
College of Engineering
Abstract
Massively Multiplayer Online Role Playing Games (MMORPGs) are a prominent genre in today's video game industry with the most popular MMORPGs generating billions of dollars in revenue and attracting millions of players. As they have grown, they have become a major target for both technological research and sociological research. In such research, it is nearly impossible to reach the same player scale from any self-made technology or sociological experiments. This greatly limits the amount of control and topics that can be explored. In an effort to make up a lacking or non-existent player-base for custom-made MMORPG research scenarios A.I. agents, impersonating human players, can be used to "bootstrap" the research scenario to reach the necessary massive number of players that define the game genre.
This thesis presents a system that makes its human players and A.I. players indistinguishable while preserving the basic characteristics of a typical MMORPG. To better achieve identical perception of human and A.I. players, our system centers around the collection, sharing, and exchange of information while limiting the means of expression and actions of players. A gameplay scenario built on the Panoptyk engine was constructed to imitate gameplay experienced in major MMORPGs. We conducted a user-study where subjects play through the scenario with a varying number of A.I. players unknown to them. Three versions of the scenario were created to assess how indistinguishable human and A.I. players were and vice versa.
We found, across 24 participants, there were 32% correct identifications, 30% incorrect identifications, and 38% answers of "I don't know". This was broken down into 20% correct identifications, 42% incorrect identifications, and 38% answers of "I don't know" for bot characters and 46% correct identifications, 16% incorrect identifications, and 38% answers of ``I don't know'' for human characters.