Quantifying Individual Player Differences

Ph.D. thesis abstract

Giel van Lankveld

 

Promotors: Prof.dr. H.J. van den Herik, Prof.dr. A.R. Arntz
Co-promotor: Dr.ir. P.H.M. Spronck

Date of defense: February 27, 2013

Computer games have existed for over 60 years and they are a popular medium of entertainment. Recently, computer games are being explored for other purposes such as education and assessment. Obviously, players of computer games vary in personality. We see these differences by looking at differences in play and by looking at the emotional and cognitive responses of players.

We aim to investigate the interaction between a player’s psychology and the game content. More knowledge about this topic is desirable for two reasons: (1) to increase our control and understanding of the experience a player has while he is playing a game and (2) to be able to adapt content to suit a player. Currently, our knowledge of the psychology at work during game-play is limited and exploration of player psychology is our main interest.

The problem statement of this thesis is: To what extent are games an appropriate means for measuring the differences between individuals based on psychological theories? To investigate the problem statement we examine incongruity theory and personality theory. We investigate the influence of these two theories on expressed emotions, on behaviour in games, and on responses on personality tests. We conduct five investigations to explore the extent to which psychological theories can explain the differences between individuals.

In Chapter 2 we present the background on the psychological concepts connected to our research and to previous research of related topics. We extensively discuss the concept of modelling from different theoretical perspectives.

After presenting the background information, Chapter 3 starts by presenting our research on RQ1: To what extent are games suitable for measuring incongruity? Incongruity theory states that players should feel boredom in easy games, pleasure in balanced games, and frustration in hard games. We investigate the relationship between the level of complexity of the player’s mental model and the emotions the same player expresses. We implement a game called Glove in which players can be confronted with a scenario of low, balanced, or high complexity (i.e., an easy, a balanced, or a hard game). From the results in Chapter 3 we may conclude that players feel frustrated when playing a hard game and that they feel pleasure when playing a balanced game. We are unable to conclude that players feel boredom when playing an easy game.

In Chapter 4 we investigate RQ2: To what extent can games be used to measure complex psychological processes such as extraversion? We investigate the process of extraversion by incorporating several extraversion experiments done in the past into a scenario programmed for the game Neverwinter Nights. In this scenario, players follow a short storyline while they perform in-game tasks imported from extraversion literature. While the players play, we record their responses on the in-game tasks. From our results we may conclude that, for 12 of the 21 recorded behaviours, correlations to extraversion or to one of the facets of extraversion are found. It is our opinion that, with additional fine-tuning, this approach can be used to measure extraversion to a larger extent. Extraversion is only one of the five personality traits in the “Big-Five” personality model.

In Chapter 5 we attempt to expand our investigation to all personality traits. There we investigate RQ3: To what extent can a data-driven personality profile be created based on game behaviour? We investigate all five traits of the “Big-Five” personality model. We wish to focus our behavioural measures on in-game behaviour, rather than on explicitly formulated replications of experiments. We consider a behavioural measure approach to be applicable in more game situations. We implement a new scenario for the game Neverwinter Nights. This scenario conforms to the scenarios found in commercial computer games. We attempt to provide the players with a broad range of possible responses to the situations encountered in the game in order to enable the free expression of the player’s personality. Because this approach is data-driven we construct 217 unpooled game variables that record the player’s movement, conversation, and general data in the game. We also formulate 43 pooled variables in order to investigate some of our a priori assumptions. We conduct correlation analyses for all game variables with the five personality traits. We also perform linear regression analyses. From our results we may conclude that personality effects for all five traits are expressed in game behaviour in both our correlation analyses and our regression analyses. Therefore, we are able to form a personality profile based on game behaviour. An approach with so many variables for these analyses runs the risk of overfitting for the number of experimental participants we used. In order to reduce the risk of overfitting, in the next chapter we focus on a theory-driven approach.

In Chapter 6 we investigate RQ4: To what extent does a theory-driven model explain personality in games? In order to investigate this question we formulate eleven behavioural criteria based on behavioural descriptions given in the “Big-Five” literature. We compute the criteria from the dataset gathered in the investigation of Chapter 5. We conducted a correlations analysis between the eleven theoretical variables and the “Big-Five” personality traits. From the results we may conclude that the neuroticism and the agreeableness variables lead to correlation with their respective personality traits. One additional conscientiousness variable correlated with the neuroticism trait. This approach shows us that the assumptions we made regarding what behaviour to expect to occur for each personality trait should be carefully tested. In the next chapter we focus on validating the results we found in this chapter and in Chapter 5.

In Chapter 7 we investigate RQ5: To what extent can our models of personality in games be validated in different games? In order to see if the results we found in the previous chapter are valid across games we implement the game behaviour variables from Chapter 5 in the starting scenario of the commercially successful game Fallout 3. After collecting the data, we analyse the dataset using both the data-driven as well as the theory-driven analyses. We may conclude that the results from the data-driven approach can be replicated to some extent but the results from the theory-driven approach are replicated less successfully.

In Chapter 8 we present a discussion on the insights that were gained during the investigations in this thesis. We may conclude that validation, by a sufficient number of participants, and control variables are points of attention when performing psychological research in a game setting. We also discuss the measure of success that we have had with our chosen approaches. We attempt to provide advice in order to avoid pitfalls and we attempt to illuminate areas of interest that may merit future research attention, such as skill, preference, intelligence, and demographics.

In Chapter 9 we present our conclusions. We provide our answers to the research questions posed in Chapter 1 and we provide our answer to the problem statement posed in that same chapter. Finally, we may conclude that games can successfully be used to quantify individual player differences based on psychological theories but that care must be taken with regards to analysis, validation, the number of participants used, and the interpretation of results.