In September 2010, eight months after the release of their space epic Mass Effect 2, developers BioWare discussed some of the strange statistics they were getting from people who were playing the game. Amidst the talk about one gamer who took 66 hours to complete the game, and the complaints over why more people didn’t play as female Shepard, an interesting nugget of information lay more or less unexplained. PC gamers were more likely to finish the optional quest for Miranda, one of the non-player characters, a quest described by BioWare themselves as having “a touchy-feely plot”. Another character – the emotionless, bloodthirsty Krogan called Grunt – was favored by XBox 360 players more.
What makes a PC gamer more likely to choose the emotional, family-issues quest over a killing spree?
The more game developers look at their own creations, the more they see unusual patterns emerging. Why do some people prefer to play as support classes? Why do some feel the need to complete every last sidequest? What makes a PC gamer more likely to choose the emotional, family-issues quest over a killing spree?
A growing group of academic researchers think they can explain these phenomena – and moreover, they think that understanding these patterns of behavior might lead to better game design in the future. At this year’s Computational Intelligence in Games conference in Seoul, South Korea, I met one such researcher who had some startling results to present about an experiment he conducted in another BioWare RPG – 2002’s Neverwinter Nights.
Giel Van Lankveld came into computer science via an unusual route – four years studying experimental psychology. His PhD supervisor, Pieter Spronck at the Netherlands’ Tilburg University , was interested in pursuing work to help model and understand how gamers behave within games, and hoped that Giel’s psychology background would bring a new angle to the research area. Understanding gamers and learning about what kind of people they are is important to game researchers, because doing it in real time allows games to become “adaptive” – redesigned to be personal to you, in ways so subtle you may not even notice.
Picture this: You’re playing the next Deus Ex game. You’re taking the stealth approach a lot, and the game notices this. In fact, over the first hour or two of the game, it’s built up a profile of what kind of gamer you are. Now it’s got enough information to confidently redesign the game to be better for you – it invents new upgrades for you to purchase that are specifically catered to stealth players. It rewrites later missions to have more hidden entrances. Not only that, but it makes the game more interesting – increasing the frequency of alarm panels and making guards more alert. You and every other gamer on the planet play the same game at heart, but the finer details are tweaked to better suit you.
Both the research world and the game industry are keen to get this technology into games, but the first step towards doing this is understanding what players are like. “Each player has a mental model of how the game works, but we can’t open up a player’s mind to look at it,” Giel told me as he explained how his research started. “We need to look at the player’s behavior and infer how accurate their model [of the game rules] is from this behavior.” So Giel began with a simple series of experiments investigating game difficulty. He wrote a program that watched participants play a simple game and then estimated how well the player understood the basic rules of the game and the strategies that arose as a result. Giel’s hypothesis was that players who had a better understanding of the game’s rules would be better at formulating strategies for defeating the enemies they faced. The data collected showed promise, but left Giel wanting more.
“The initial work was successful, but limited,” Giel explains. “We wanted to get beyond just difficulty adjustment and into [game] content in general.” The problem was that modeling player understanding was fine for simple things like making levels easier if the player wasn’t particularly skilled. But in order to adjust all the content in a game – from the story to the graphical style – researchers would need to understand the player on a more fundamental level. This led Giel to wonder about the player’s psychology in general, even their entire personality. “If we can unravel the relationship between personality and game content we can start offering content that fits with the player.” The question was, how do you tell a gamer’s personality from watching him hit goblins with a pointy stick?
The question was, how do you tell a gamer’s personality from watching him hit goblins with a pointy stick?
You’ve probably taken a personality test or two in your time – big long lists of questions about who you are, what you’re like and what you feel about this and that. These tests, more often than not, are designed to break your personality down into five key categories, known to the psychological community by the acronym OCEAN – Openness, Conscientiousness, Extraversion, Agreeableness and Neuroticism. Giel’s next step would be to try and obtain this information without asking a single question of the player.
To do this, Giel designed a scenario for Neverwinter Nights. In the scenario, players are guided through a short dream sequence serving as a turtorial before they’re let loose in a small quest line – a town’s water supply has been poisoned and the player must find out how and stop it. The scenario features sidequests, some combat and a collection of characters to talk to – similar to most of the Neverwinter Nights modules available online. The only difference here is that as the player progresses, the module records data for Giel.
Before playing the scenario, each of Giel’s 44 participants – a mix of both gamers and non-gamers – took a standard personality questionnaire, so that Giel could compare his game data to the person’s real-world personality. Then, after Giel had more than forty hours of gameplay data and another thirty hours of questionnaires to sift through, he began the arduous task of searching for connections between activity in the game and personality attributes in real life. Expecting one or two moderately strong relations, Giel was surprised to find many significant relationships between personality traits and in-game behavior.
The results in his paper, Games as Personality Profiling Tools, highlight the correlations between certain in-game actions and high or low scores in personality traits. The results are fascinating. For example, participants in the experiment who exhibited high Agreeableness in the personality interview – indicating that they are friendly and compassionate – talked less with aggressive or rude characters in the scenario, and were more likely to warn others about the poisoned water source once they discovered it. Those with high Neuroticism scores – associated with nervousness or sensitivity – took far longer to finish the scenario, many exceeding an hour of play time. Conscientious players – a factor that relates to how organized or efficient someone is – explored many of the game’s longer conversations to exhaustion.
In fact, Giel analyzed found statistically significant correlations in over 180 of out of the 275 variables he recorded in-game. These included conversation triggers, quest choices and progress, time taken to perform tasks and dozens more. Interestingly, although many of the correlations revealed fascinating connections between real-life personality and in-game behavior, there were still things that puzzled Giel. When we spoke about the results, he highlighted the Extraversion trait – associated with outgoingness, energetic tendencies and contact with other people – and how it correlated inside the game.
Extraverts also seemed to delight in “risky” conversations – becoming argumentative with people in the village tavern, for instance.
“What surprised me a lot was that extraversion is not super dominant in our data. All across the literature you see that extraversion in ‘normal’ life is omnipresent. It determines how much people talk, smile and interact with others. It determines what type of activities people like. It’s virtually everywhere.” But even in this case, where he found fewer correlations than expected, Extraversion still features in Giel’s results. High scorers seemed to spend the most time in the dream sequence at the start of the scenario, which Giel attributed to being easily stimulated by the strange light effects and unusual situation. Extraverts also seemed to delight in “risky” conversations – becoming argumentative with people in the village tavern, for instance.
For the future, Giel will be focusing on finishing his PhD, and then hopefully returning to the question that sparked off his research initially – whether this work on personality modeling can help shape the next generation of videogames. “Our main motivation when starting all this was to find a way to adapt game content based on personality,” Giel explains. “But we found out that it takes way more effort to actually do this research than we expected. Collecting data from humans takes a lot of time.”
Giel is already planning a new series of experiments to attempt to find a link between a player’s personality and their preference for different sorts of content. Early experiments have shown mixed results, but he remains upbeat. “Videogames are an untapped resource. We know very little about the relationship between game behavior and the cognitive processes behind it.” And where little is known, there is much research to be done.
Whether we’ll be seeing personalized content any time soon is uncertain – as with many research projects, real-world application remains an open question. This year’s Computational Intelligence and Games conference was heavily sponsored by MMO developers NCSoft, who also attended a number of talks including Giel’s. Giel’s passion for his work, however, makes it clear that getting this research into games is not the main concern of his research. Regardless of whether this technology worms its ways onto our hard drives, researchers like Giel remain resolute in uncovering the underlying psychology of the average gamer, and their work promises many more surprising results in future.
Michael Cook is a PhD researcher in automated creative design, including game design. He would like you to lie down and tell him about your father.