以機器學習技術建構玩家人格特質模型

Building Player’s Personality Trait Model Using Machine Learning Techniques

劉庭愷(Ting-Kai Liu)

指導教授 ： 孫春在

國立交通大學/資訊學院/資訊科學與工程研究所/碩士(2020年)

摘要

遊戲於現今的時代十分普及，逐漸增長的玩家數量蘊含了大量的資料，透過對玩家遊戲行為的觀察，能了解玩家的個人特質，這樣的觀察如果能被歸納整理，對個人特質的研究就能有更進一步的發展。隨著玩家投入遊戲，虛擬世界中提供的環境讓玩家能展現豐富多樣的行為，透過這樣的方式來分析玩家個人特質會比傳統的問答測驗方式來的更好。本研究主題在探討是否能透過分析玩家於遊戲中的行為紀錄，結合機器學習技術，取代傳統問答測驗方式來分析玩家的人格特質。
本研究分析了七十三名玩家的遊戲行為資料，包含玩家的滑鼠軌跡與眼動軌跡，並使用台灣大專院校普遍使用的賴氏人格測驗，分析玩家的遊戲行為與人格測驗結果間的關聯性，將玩家遊戲行為作為預測模型的輸入資料，用來預測玩家的人格測驗結果。
研究結果顯示人格測驗結果能用預測模型進行預測，對人格特質量表來說不需要十分精確的數字，僅需要知道受測者每項特質的傾向，而在本研究中雖然只使用了簡單的回歸模型做預測，就足以使用玩家於遊戲中的眼動行為和滑鼠行為預測玩家人格特質的傾向。其中本研究使用的遊戲對人格特質中活動性（G）、自卑感（I）、緊張性（ST）、焦慮性（AN）的預測表現較好，且認為不同的遊戲類型適合不同的人格特質展現。

關鍵字

機器學習 ； 人格 ； 人格特徵 ； 眼動 ； 滑鼠 ； 遊戲行為

並列摘要

Games have become a widespread entertainment feature nowadays. Therefore, the increase of players can provide massive quantitative information. We rely upon observations of gaming behavior to understand player’s personality traits, which would be useful in analyzing human personality. A virtual world can provide a rich environment for player immersed in the game world to express a variety of different behaviors. This kind of assessment through gameplay is better than traditional personality assessment. This research aims to find patterns in the players’ behavioral data, using machine learning techniques, to create a model that can replace traditional personality assessment.
In this study, 73 players’ gameplay behaviors were analyzed, including mouse movement and eye movement. They were asked to fill out Lai personality assessment, which is a popular used psychological test in the universities in Taiwan. We then use the gameplay behaviors to predict players’ personality assessment result.
The result of our experiment shows that personality assessment result can be predicted by machine learning model. Although we only use simple regression model for prediction, it is enough for us to predict player’s personality traits tendency by mouse movement and eye movement. Among all the personality traits prediction, gregariousness, inferiority, stress, and anxiety performed well. Furthermore, to show particular personality, we need different types of games.

並列關鍵字

Machine Learning ； Personality ； Personality Trait ； Eye Movement ； Mouse Movement ； In-Game Behavior

參考文獻

[1] J. H. Kim, D. V. Gunn, E. Schuh, B. Phillips, R. J. Pagulayan, and D. Wixon, “Tracking real-time user experience (TRUE): a comprehensive instrumentation solution for complex systems,” in Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, Florence, Italy, 2008, pp. 443–452.

[2] R. Ganellen, “Assessing Normal and Abnormal Personality Functioning: Strengths and Weaknesses of Self-Report, Observer, and Performance-Based Methods,” Journal of personality assessment, vol. 89, pp. 30-40, 09/01, 2007.

[3] S. Tekofsky, P. Spronck, A. Plaat, A. Plaat@uvt, J. Nl, H. Herik, J. Broersen, J. Broersen@uu, and Nl, PsyOps: Personality Assessment Through Gaming Behavior, 2013.

[4] P. R. Matthew Stewart. "The Actual Difference Between Statistics and Machine Learning," Apr. 2, 2020; https://towardsdatascience.com/the-actual-difference-between-statistics-and-machinelearning-64b49f07ea3

[5] D. Hansen, and Q. Ji, “In the Eye of the Beholder: A Survey of Models for Eyes and Gaze,” IEEE transactions on pattern analysis and machine intelligence, vol. 32, pp. 478-500, 03/01, 2010.