Commercial Video Games as Preparation for Future Learning
Commercial Video Games as Prepara1on for Future Learning Dylan Arena Stanford University June 2012Tuesday, June 26, 12Hi! OK, I’m gonna get started. I’ll be talking today about the poten1al of using commercial video games as prepara1on for future learning.
BackgroundTuesday, June 26, 12First I’ll set the stage a little bit…
Tuesday, June 26, 12We in the GLS community claim that gameplay can be great for learning, and by that we tend to mean two things:
Tuesday, June 26, 12First, that we can design learning-‐based games (like Quest Atlan1s) to support “schoolish” learning…
Tuesday, June 26, 12…and second, that we can observe other interes1ng kinds of learning even in commercial, oﬀ-‐the-‐shelf games (like World of WarcraY).
Tuesday, June 26, 12A third claim—that simply playing commercial games recrea1onally…
Tuesday, June 26, 12—seems a bit silly at ﬁrst, if only because most schoolish tests focus on retrieval of facts that most commercial games aren’t designed to teach.
Tuesday, June 26, 12But if we broaden our view, it’s a bit less silly.
TheoryTuesday, June 26, 12Here’s a thought experiment to help make clear why.
Tuesday, June 26, 12Imagine spli]ng a class in half: half get to explore a forest (turning over rocks, looking inside ro_en logs, poking ant hills) and half don’t (they just stand quietly facing the wall).
Forest ecosystem testTuesday, June 26, 12Then bring both groups back into class and give ‘em a tradi1onal mul1ple-‐choice test about forest ecosystems. You’d expect no diﬀerence, right? It’s unlikely that just exploring the forest would give kids the kind of knowledge they’d need to pick correct answers on a mul1ple-‐choice test.
Forest ecosystem test Forest Ecosystem LectureTuesday, June 26, 12But imagine then giving a LECTURE about forest ecosystems—ground cover, canopy, decay and new growth, etc. It’s possible that the kids who had just been out exploring a forest might engage more with that lecture and hence learn more.
Forest ecosystem test Forest Ecosystem Lecture Forest ecosystem testTuesday, June 26, 12If you then test everyone again aYer the lecture, you might observe some previously hidden beneﬁts of the forest ﬁeld trip.The point is that what we bring into a learning situa1on (like a lecture) is obviously very important. But it can be hard to measure what we bring in, especially when it isn’t stable, well-‐structured factual knowledge.
Prepara1on for Future Learning (PFL) [Bransford & Schwartz, 1999] Bad (or no) Good Experience Experience Sequestered-‐Problem-‐Solving test Future Learning Prepara1on-‐for-‐Future-‐Learning testTuesday, June 26, 12That’s where the “Prepara1on for Future Learning” part of my 1tle comes in. PFL is an assessment framework designed to measure inchoate forms of prior knowledge that tradi1onal (or “sequestered-‐problem-‐solving”) tests miss.This slide shows a generalized diagram of the forest-‐test-‐lecture-‐test scenario I just described: some learners have a good (which is to say, learning-‐relevant) experience, and others don’t. On a Sequestered-‐Problem-‐Solving test, they look about the same. But if you then provide a learning opportunity that is designed to mold that prior experience into a formal knowledge structure and test ‘em again, you can detect the beneﬁts (or lack thereof) of the experience.
Study DesignTuesday, June 26, 12In a moment I’ll talk about how I used the PFL framework, but first, a point about mystudy-design goals.Reports on the state of the field, like the 2011 National Research Council report,describe the evidence for games supporting schoolish learning as “emerging”,“inconclusive”, and “very limited”, with “gaps and weaknesses” that “make it difficultto...demonstrate their effectiveness...”These statements reflect the fact that many stakeholders want from the GLS communitysomething like an FDA study: a randomized field trial with an intention-to-treat analysisand very traditional, schoolish operationalizations of learning. So I decided to try to runone.
Study DesignTuesday, June 26, 12Here’s a first pass at the study design, stripped down so you can see the parallels with the forest-field trip example (more details will follow).I randomly assigned community-college students to three conditions: play Civilization 4, play Call of Duty 2, or play no game. (I just gave gameplay participants the games they’d been assigned and told ‘em to play at home however they normally play, for at least 15 hours over the course of about 5 weeks.)
Study DesignTuesday, June 26, 12Then I had all participants come in and take a 16-item multiple-choice test about World War II history.
Study DesignTuesday, June 26, 12Then I had them watch a 20-minute narrated-slideshow lecture about World War II history.
Study DesignTuesday, June 26, 12Then I gave ‘em another multiple-choice test about World War II history (this time 36 items).
Study DesignTuesday, June 26, 12So that’s the basic study design. Here I’ll fill in a few more details.
Study DesignTuesday, June 26, 12First, the participants were 102 local community-college students (Control: n = 33; CoD2: n = 34; Civ4: n = 35; 16-42 yrs, median 20 yrs; 64% female) whom I compensated with course credit and (if they played for the full 15 hours) pay; all had completed a huge demographic questionnaire (roughly 280 questions) as part of their research-participation program; and the way I explained the study was that everyone would get a free game and (possibly) a gift card, with the only differences being *which* game and *when* the gameplay would occur (before or after the in-person session); this way Control participants wouldn’t feel shortchanged.
Study DesignTuesday, June 26, 12The way I verified gameplay was by collecting and analyzing participants’ save-game files (auto-generated by games so players can pick up where they left off).
Study DesignTuesday, June 26, 12All players whose save-game files showed evidence of at least 15 hours of gameplay were compensated with $75 gift cards (45 people earned ‘em: 11 Control, 15 CoD2, 19 Civ).(And you can see here that the “control” participants did receive a game to play for 15+ hours and got compensated if they did so.)
MaterialsTuesday, June 26, 12Now I’ll describe the games, lecture, and measures.
The GamesTuesday, June 26, 12I chose these games because they were (a) both really popular with players and critics when they were released in 2005; (b) from successful franchises of games; (c) and old enough to be playable on any modern computer but still new enough to seem “cool”.
The Games: Civilization IVTuesday, June 26, 12Civ4 is a turn-based-strategy game.
The Games: Civilization IVTuesday, June 26, 12You play as the immortal, autocratic ruler of a civilization, and your task is to guide your people through roughly 6000 years of history by making lots of choices.
The Games: Civilization IVTuesday, June 26, 12You’re plopped down on some arbitrary Earth-like world, and you settle cities, build infrastructure, engage with other civilizations through diplomacy and/or warfare, and create wonders based on those in Earth’s history (like the Parthenon, or Rock ‘n’ Roll, or the Manhattan Project).
The Games: Civilization IVTuesday, June 26, 12By the end of the game, you’ll have built a bunch of cities, fought some wars, and made a ton of choices.
The Games: Call of Duty 2Tuesday, June 26, 12CoD2 is a first-person-shooter game.
The Games: Call of Duty 2Tuesday, June 26, 12You play as a lowly soldier: a Soviet peasant repelling the German invasion; then later a Brit in the North African campaign; and finally an American in the invasion of France, ending the game by crossing the Rhine into Germany.
The Games: Call of Duty 2Tuesday, June 26, 12In contrast to Civ4, CoD2 is a real-time game in which you are required to navigate a 3D environment and shoot things (like teddy bears).
The Games: Call of Duty 2Tuesday, June 26, 12You play as part of a small squad of soldiers overcoming various obstacles to reach the next objective, which is marked as a gold star on your map (lower left). CoD2 gameplay is fast, twitchy, and visceral, not much time for thoughtful reflection—you just shoot whatever threatens you and move toward the next gold star. But it’s all happening in the historical context of WWII theaters of war.
The LectureTuesday, June 26, 12The lecture covered WWII from the initial troubles in Asia in the 1930s to the dropping of the atomic bombs. My primary resource for the lecture was a SparkNotes guide (like Cliff’s Notes: high-schoolers might use it to study for their history tests). I wrote the lecture to cover all of WWII but also to focus on two sets of themes, corresponding to the gameplay experiences that I predicted the two games would produce.
The LectureTuesday, June 26, 12I had hoped that (because of their gameplay experiences) Civ4 players would engage more with the Nations themes and CoD2 players would engage more with the Battles themes.
The Tests [SparkNotes] Pre2.1 From the perspective of Western leaders, Stalin’s actions as leader of the Union of Soviet Socialist Republics reflected an emphasis on which of the following concepts? individualism freedom human dignity aggression [NAEP] Post2.31 When the United States entered the Second World War, one of its allies was Germany Japan the Soviet Union Italy [CST] Post2.36 The purpose of the Manhattan Project was to provide economic aid to Latin American countries develop atomic weapons for the U.S. military bring about an end to poverty in U.S. urban areas offer assistance to relocated European refugeesTuesday, June 26, 12I built the pre- and post-lecture tests using traditional multiple-choice items that I pulled from three sets of standardized tests: a quiz from the back of the SparkNotes guide I mentioned; the National Assessment of Educational Progress; and the California Standards Tests. Here are three items.
Open-Ended Questions After post-lecture test, two scenarios not mentioned in the lectureTuesday, June 26, 12So these traditional tests were my nod to the conservative folks out there who think that learning is factual retrieval. But we at GLS know better! So in addition to these traditional tests, I also included two sets of open-ended questions that described scenarios not mentioned in the lecture.
Open-Ended Questions After post-lecture test, two scenarios not mentioned in the lecture One for Nations, one for BattlesTuesday, June 26, 12One of these two scenarios was designed to pick up on a focus on the Nations themes I had tried to build into the lecture (and hence to favor Civ4 players), while the other was designed to pick up on a Battles focus (and hence favor the CoD2 players).
Open-Ended Questions After post-lecture test, two scenarios not mentioned in the lecture One for Nations, one for Battles Two questions per scenario: What’s going on? What would you want to ask to learn more?Tuesday, June 26, 12Participants were asked what they thought was going on in each scenario and, more importantly, what questions they’d want to ask to learn more.
Open-Ended Questions (Nations focus) In 1940, in Mers-el-Kebir, Algeria, commanders of some British ships spoke with commanders of some French ships, and then the British ships fired on the French ships, sinking the ships and killing over 1,200 French sailors. Why do you think this might have happened? (Feel free to guess.) What questions would you ask to figure out why this happened? (Dont just say, "I would ask why this happened." Thats too easy. Think about what kinds of facts about the situation you would want to know.)Tuesday, June 26, 12Here’s the first scenario (the Nations-focus one) and its two questions…
Open-Ended Questions (Battles focus) On June 6, 1944, an American Ranger battalion landed on the beach at the foot of the cliffs of Pointe du Hoc, in France. They then climbed those cliffs under fire from the Germans to destroy a set of large artillery guns. Why do you think the Americans wanted to destroy the guns? (Dont just say, "To stop the Germans from using them." Be specific. Think about where the Germans might have wanted to use the guns.) What questions would you ask to figure out why this happened? (Dont just say, "I would ask why this happened." Thats too easy. Think about what kinds of facts about the situation you would want to know.)Tuesday, June 26, 12…and here’s the Battles-focus scenario and questions.
ResultsTuesday, June 26, 12My analysis protocol for the multiple-choice tests was to fit ANCOVA models using test scores as the outcome variables. To choose predictors, I created a candidate list of nine that I had reason to believe might be useful...
Candidate predictors for ANCOVA models • Gameplay conditionTuesday, June 26, 12Gameplay condition (operationalized differently by analysis)
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collectionTuesday, June 26, 12quarter of data collection (to account for cohort effects)
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collection • GenderTuesday, June 26, 12gender
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collection • Gender • AgeTuesday, June 26, 12age
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collection • Gender • Age • Prior gameplay historyTuesday, June 26, 12prior-gameplay history (4-level ordinal from “never” to “> 6 times”)
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collection • Gender • Age • Prior gameplay history • English proficiency levelTuesday, June 26, 12English proficiency (lots of non-native speakers in my sample)
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collection • Gender • Age • Prior gameplay history • English proficiency level • Prior social-studies interestTuesday, June 26, 12prior social-studies interest (5-level Likert)
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collection • Gender • Age • Prior gameplay history • English proficiency level • Prior social-studies interest • Enjoyment of the assigned gameTuesday, June 26, 12enjoyment of the assigned game (obviously only relevant for gameplay participants)
Candidate predictors for ANCOVA models • Gameplay condition • Quarter of data collection • Gender • Age • Prior gameplay history • English proficiency level • Prior social-studies interest • Enjoyment of the assigned game • Pre-lecture-test scores (for post-lecture-test ANCOVA)Tuesday, June 26, 12and pre-lecture-test scores (obviously only for the post-lecture-test ANCOVA).I then fed this candidate set of predictors into an all-possible-subsets selection procedure, which examines all combinations of the predictor set to find the model with the highest adj-R^2. To avoid capitalizing on spurious patterns in the data set, I also constrained it to include only models all of whose predictors were at least marginally significant. I call the resulting model the “parsimonious” model.
Parsimonious model for pre-lecture-test scores Source df SSTypeIII F η2 p English proficiency 1 70.86 13.32 .12 .00042** Residuals 100 532.16 R2adj = .11, F(1, 100) = 13.32, p = .00042Tuesday, June 26, 12The only predictor of pre-lecture-test scores was English proficiency. This is the SPS test from our PFL model. NOTE: If this were all we did to test the learning benefits of recreational commercial gameplay, we’d find no benefit. But on the post-lecture-test...
Parsimonious model for post-lecture-test scores Source df SSTypeIII F η2 p Received a game 1 88.97 4.05 .026 .047* Quarter 2 125.40 2.85 .036 .063. Age 1 123.55 5.63 .036 .020* English proficiency 1 104.52 4.76 .030 .032* Prior SS interest 4 334.12 3.80 .096 .0068** Game enjoyment 4 239.07 2.72 .069 .035* Pre-lecture test 1 145.08 6.60 .042 .012* Residuals 87 1910.95 R2adj = .36, F(14, 87) = 5.06, p < .0001Tuesday, June 26, 12...all of a sudden a lot is going on. The key points for this talk are (a) many things are involved in how players will learn from gaming experiences, and (b) gameplay participants significantly outscored control participants—i.e., they learned more from the lecture.In fact, gameplay participants scored about 6% higher on the post-lecture test than did control participants (without considering covariates, just a straight means comparison). In the language of school, that translates to over four percentage points on an exam (74.3% for control participants and 78.6% for gameplay participants), or nearly half of a letter grade. Cohen’s d = .27, which is substantial for a randomized field trial.
Responses to open-ended questions Nations focus Control CoD2 Civ4 Battles focus Control CoD2 Civ4 No 22 21 15 No 22 16 26 Yes 11 13 20 Yes 11 18 9 Fisher’s exact test: p = .058 Fisher’s exact test: p = .030Tuesday, June 26, 12The other outcome measure, remember, was participants’ responses to open-ended questions about novel WWII scenarios. It turned out that participants’ gameplay experiences affected the focus of their responses, with Civ4 participants adopting a more global “Nations” focus and CoD2 participants adopting a more local “Battles” focus.(The specific operationalization of my coding scheme for this scenario was to code participants as having a “Nations” focus if and only if (a) the participant’s questions mentioned Resources (including territory), Empires (including colonies), Defenses (including enemies), or Alliances (including treaties), or (b) the participant wrote of the actors as being the nations themselves (e.g., Britain, France) rather than agents of those nationalities (e.g., British commanders, French ships).The operationalization of my coding scheme for this scenario was to code participants as having a “Battles” focus if and only if the participant’s questions mention (a) Weaponry (including capabilities of particular weapons), Terrain (including avenues of ingress for the engagement), Communication (but not including prior intelligence about the engagement), or Objectives (but not including consequences of the engagement) or (b) such tactical elements as the time course of the engagement, casualties, or troop size. (CoD2 participants’ responses tended to reflect the in-the- moment viewpoint of a soldier anticipating climbing those cliffs to engage an enemy.))
DiscussionTuesday, June 26, 12I’ve got three basic takeaways from these results, and then three suggestions for various stakeholders.
Summary of Findings • Playing enjoyable video games at home can help students learn in school(ish settings)Tuesday, June 26, 12First, the results of this study support the claim that playing enjoyable video games at home can help both male and female students learn in school, if the formal instruction leverages the students’ gameplay experiences. (The strong predictive effect of prior social-studies interest shows the importance of also leveraging students’ interests.)
Summary of Findings • Playing enjoyable video games at home can help students learn in school(ish settings) • Different game experiences lend themselves to different types of instructionTuesday, June 26, 12Second, the results from the open-ended questions underscore the notion that different games will offer different types of experiences that prepare players preferentially for different topics of formal instruction.
Summary of Findings • Playing enjoyable video games at home can help students learn in school(ish settings) • Different game experiences lend themselves to different types of instruction • Gameplay can influence both retention of facts and choices about what to learnTuesday, June 26, 12And third, the open-ended-question results further suggest that these gameplay experiences can improve not only retention of facts presented by direct instruction but also students’ choices about what to learn.
Considerations for Practice • Curriculum Designers: From task analysis of games to curriculumTuesday, June 26, 12Now, my three recommendations for folks who want to cash these results out in some way.First, for folks who want to build curricula to leverage gameplay experiences, I’d suggest a careful task analysis of gameplay to determine the relevant properties of its experiences (e.g., for Civ4 it was thinking as a nation; for CoD2 it was probably as simple as just being exposed to the historical context of WWII).
Considerations for Practice • Curriculum Designers: From task analysis of games to curriculum • Educators: Gameplay doesn’t have to be wasted timeTuesday, June 26, 12For educators who have to deal with their students playing games for hours each week, I’d suggest that they recognize that gameplay is pervasive and powerful and that they embrace it (by tying the compelling experiences found in games with the powerful explanatory structures found in the standard curriculum).
Considerations for Practice • Curriculum Designers: From task analysis of games to curriculum • Educators: Gameplay doesn’t have to be wasted time • Game Designers: You can be chickens, not pigsTuesday, June 26, 12And for commercial game designers—who know how hard it is to make a good game, let alone a good learning game, and have therefore steered clear of the educational game space—I’d say that this study suggests that they needn’t try to cram all of the curricular content into the game itself. Instead, they can continue to let the game do what it does best (provide great experiences) with perhaps some small tweaks here and there to better serve as foundations upon which educators might build.(from a joke about eggs/bacon for breakfast: the chicken is interested, but the pig is committed)
Acknowledgement Financial support for this dissertation was provided by a SUSE Dissertation Support Grant and a Gerald J. Lieberman FellowshipTuesday, June 26, 12
Tuesday, June 26, 12And that’s it—thanks very much for your time!Questions?