Becoming a Gamer: Cognitive Effects of Real-Time Strategy Gaming Brian Glass The University of Texas at Austin, 2011 Overview • Real-time strategy (RTS) gaming • Why study gaming? • Blitzy results from 40-hour training with novices • Novel modeling technique Real-Time Strategy Games Games Video Games Action Most Existing Research e.g., First-person shooter Adventure Simulation Strategy Real-Time Strategy Games Games Video Games Action Real-Time Strategy Adventure Simulation • God-like view / command • Rapid, fluid action • Small window into larger world (literal spotlight) • Resource management • War themes, many “fronts” Strategy Why video games? 1. Popular 2. Lead to generalized learning 3. We can control them Why video games? Popular • 72% of Americans reported playing video games in 2008 • Females are fastest growing demographic • In 2010, 40% of gamers were female • Of all gamers, 33% were adult women > 20% who were males under 17 • Striatal dopamine during video game was on par with methamphetamine injection (Koepp, et al., 1998) • DSM-V committee found insufficient evidence for video game addiction (as of 2007) Entertainment Software Association, 2010 Essential Facts Guide Why video games? Lead to generalized learning • Transfer to new tasks and new contexts • Prior work demonstrates that long-term training in attention-based tasks can lead to performance enhancement • Most training procedures tap specific skill specific system Green & Bavelier (2008) Existing Research: Action Games Players vs. Non-Players • • Self-report as avid action gamers (5+ hours per week) Tested on various speeded-choice RT tasks Dye, Green, & Bavelier (2009) Existing Research: Action Games Training Non-Players • Pretest – 50 hours of play over 9 weeks – Posttest Dye, Green, & Bavelier (2009) Existing Research: Action Games Training Non-Players • Multiple-object tracking task • Useful field of view task • Attentional blink (10 Hz) • Crowding tasks (identification among tightly packed distractors) • Mental rotation • Piloting training (flight simulator) Feng, Spence & Pratt (2007) Dye, Green, & Bavelier (2009) Myers, et al. (2000) StarCraft: “Gold Standard” RTS Video Attention and StarCraft Does RTS gameplay have more of an attentional component? Strains on attention: Multi-agent system Resource management Multiple “fronts” Attention and StarCraft StarCraft • Alerting • Sensitivity to incoming stimuli Sensitivity to enemy units which may appear on the screen • Orienting • Selection of information from sensory input Selecting which game elements to attend to • Executive • Monitoring and resolving conflicts Determining priority of which enemy units to attack (Posner & Rothbart, 2007) Training Novices 3. We can control them 1. Control Game (The Sims) 2. Low-Attention Starcraft 3. High-Attention Starcraft Used in: Dye, Green, & Bavelier, 2009 Training Novices High-Attention (Two Bases) Low-Attention (One Base) Training Novices Intake / Screening Pretest 20 Hours of Gaming Midtest 1. Control (The Sims) 2. Low-Attention StarCraft 3. High-Attention StarCraft 20 Hours of Gaming Posttest Training Novices Battery of cognitive / perceptual tasks – Expect Change? Stroop – Directed attention Task Switching Multitask Switching Attentional Network Test (ANT) Visual Search Filtering Digit Span – Working Memory Operating Span (Ospan) Balloon Analogue Risk Taking Task Multimedia Multitasking Index Yes Maybe No Pretest only, possible covariate Training Novices Attentional Network Test (ANT) Fan, McCandliss, Fossella, Flombaum, Posner (2005) Training Novices Attentional Network Test (ANT) ANT, Time3-Time1 0.30 100.0 0.25 50.0 0.20 0.0 RT Accuracy ANT, Time3-Time1 0.15 -50.0 0.10 -100.0 0.05 -150.0 0.00 -200.0 SC-Low SC-High SC Sims SC-Low SC-High SC All groups faster, but StarCraft-High group more accurate Sims Training Novices Multitask Switching 0 0 3 Switching between locations, remembering states 4 Odd? Odd? 1 7 2 6 Odd? 8 Training Novices Multitask Switching Switching between locations, remembering states Multitask Switching, Time3-Time1 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 RT - Switch Trials ACC - Switch Trials Multitask Switching, Time3-Time1 SC-Low SC-High SC Sims 400.0 200.0 0.0 -200.0 -400.0 -600.0 -800.0 -1000.0 -1200.0 * SCLow SCHigh SC Sims Training Novices Task Switching 4 3 Even / Odd? 5 1 p u Vowel / Cons. j o E/O E/O E/O V/C Switch V/C E/O E/O V/C V/C V/C V/C E/O V/C Switch Switch Switch Switch Ophir, Nass, Wagner (2009) Training Novices Task Switching E/O E/O E/O V/C V/C Switch Task Switching, Time3-Time1 Task Switching, Time3-Time1 0.10 0.08 0.04 RT Accuracy 0.06 0.02 0.00 -0.02 -0.04 -0.06 SC-Low SCHigh Switches SC Sims Non-Switches 0 -50 -100 -150 -200 -250 -300 -350 -400 -450 -500 SC-Low SCHigh Switches SC Sims Non-Switches Ophir, Nass, Wagner (2009) Training Novices Filtering Information Filtering, Time3-Time1 Information Filtering, Time3-Time1 0.150 40 20 Reaction Time Accuracy 0.100 0.050 0.000 -0.050 0 -20 -40 -60 -80 -0.100 -100 -0.150 -120 SCLow SCHigh SC Sims SC-Low SCHigh SC Sims Ophir, Nass, Wagner (2009) Training Novices Balloon Analog Risk Taking (BART) • Sensation seeking • Impulsivity • Constraint deficiencies • Self-reported: • Addictive Risk Bx • Health Risk Bx • Safety Risk Bx Lejuez, et al. 2002 Training Novices Balloon Analog Risk Taking (BART) 0.12 0.10 0.08 0.06 0.04 0.02 0.00 -0.02 -0.04 Riskier Accept Length BART, Time2-Time1 SC-Low SCHigh SC Sims Lejuez, et al. 2002 Results Summary Battery of cognitive / perceptual tasks Stroop – Directed attention Task Switching* Sims SC-Lo SC-Hi + + -A / -RT +A / -RT +A / -RT Multitask Switching* -RT Attentional Network Test (ANT)* -RT -RT +A / -RT Visual Search – Target/Distractors H<E -RT H<E -RT H~E -RT Filtering -RT -A / -RT +A / -RT Digit Span – Working Memory* + + Operating Span (Ospan)* - + +Risk ++Risk Risk Balloon Analogue Risk Taking* *NHST p<0.05 3 2 Continuing Analyses StarCraft Measures • Games won Duration Final difficulty levels • Feature Modeling 70+ features are being captured during game play Feature Modeling with RADAR Feature Recording for Each Unit ID unittype isselected isabuilding unit ID type of unit is selected by the player is a building isidle idle = 1, not idle = 0, being built = 2, newly built = 3 isloaded disttocc hasgas hasminerals isrepairinganother loaded into a transporter distance to nearest command center scv carrying gas scv carrying minerals is repairing another unit tickssincelastunderattack milliseconds since last under attack prophealth killcount velocity healedbymedic blindedbymedic targetDist islifted cntfriendly cntenemy distfriendly distenemy cntfriendlybldg cntenemybldg distfriendlybldg distenemybldg proportion health remaining other units killed movement velocity being healed by a medic being blinded by a medic distance to target is lifted (building) friendly units in view enemy units in view distance to nearest friendly unit distance to nearest enemy unit # friendly buildings in view # enemy buildings in view distance to closest friendly building distance to closest enemy building cntfriendlyunderattack # friendly units under attack in view cntenemyunderattack # enemy units under attack in view distfriendlyunderattack distenemyunderattack cntfriendlybldgunderattack propvisible proportion of possible tiles in view which are unobstructed and viewable propbuildable proportion of possible tiles in view on which buildings can be built ispatrolling istraining remaintraintime isresearching remainresearchtime screenMoved_int mmDragged_int is patrolling is being trained remaining time to be trained is a building which is researching remaining time to research has the screen moved since last recording has the minimap been dragged since last recording sX sY x y critterinview angle xdivy ydivx mousedist distscreentop distscreenleft distscreencorner cntfriendlyunderattackodd screen x position on map screen y position on map x position y position critter in view angle of rotation x position divided by y position y position divided by x position distance from the mouse cursor distance from unit to top of screen distance from unit to left of screen distance from unit to screen corner # friendly units under attack is odd cntenemyunderattackodd # enemy units under attack is odd cntfriendlybldgunderattackodd # friendly buildings under attack is odd distance to closest friendly unit under attack cntenemybldgunderattackodd xodd yodd angleodd xgreatery # enemy buildings under attack is odd x position is odd y position is odd angle of rotation is odd x position is greater than y position distance to closest enemy unit under attack cntfriendlyunderattackmult3 count of friendly units under attack is multiple of 3 # friendly buildings under attack in view cntenemyunderattackmult3 count of enemy units under attack is multiple of 3 cntenemybldgunderattack # enemy buildings under attack cntfriendlybldgunderattackmult3 # friendly buildings under attack is multiple of 3 distfriendlybldgunderattack distance to clsoest friendly building under attack distenemybldgunderattack distance to closest enemy building under attack cntgas cntmineral distgas distmineral # of geysers in view # of mineral patches in view distance to closest gas distance to closest minerals cntenemybldgunderattackmult3 xmult3 ymult3 anglemult3 random1 # enemy buildings under attack is multiple of 3 x position is a multiple of 3 y position is a multiple of 3 angle of rotation is a multiple of 3 random number Feature Modeling with RADAR Feature Recording for Each Unit unittype isselected isabuilding isidle isloaded disttocc hasgas hasminerals isrepairinganother tickssincelastunderattack prophealth velocity healedbymedic distenemy cntfriendlybldg cntenemybldg distfriendlybldg cntenemybldgunderattack distfriendlybldgunderattack cntgas cntmineral propvisible yodd xgreatery cntfriendlyunderattackmult3 cntenemyunderattackmult3 ymult3 anglemult3 type of unit is selected by the player is a building idle = 1, not idle = 0, being built = 2, newly built = 3 loaded into a transporter distance to nearest command center scv carrying gas scv carrying minerals is repairing another unit milliseconds since last under attack proportion health remaining movement velocity being healed by a medic distance to nearest enemy unit # friendly buildings in view # enemy buildings in view distance to closest friendly building # enemy buildings under attack distance to clsoest friendly building under attack # of geysers in view # of mineral patches in view proportion of possible tiles in view which are unobstructed and viewable y position is odd x position is greater than y position count of friendly units under attack is multiple of 3 count of enemy units under attack is multiple of 3 y position is a multiple of 3 angle of rotation is a multiple of 3 Feature Modeling with RADAR Response Adaptive Display Anticipates Request (RADAR) Feature Modeling with RADAR • StarCraft – Correctly predicts how likely a unit is to be selected – 75% of the time. Compare Novice vs. Expert Cross-validate over games to measure consistency of game play Summary • RTS gaming represents unique opportunity for long-term cognitive training • Train 36 novices over three conditions • High-Attention RTS • Low-Attention RTS • Control (The Sims) • StarCraft vs. Control (Sims) • When accuracy is high, RT improvement • When accuracy is moderate, accuracy + RT improvement • However, Control faster at location switching task •StarCraft-High vs. StarCraft-Low • StarCraft-Low group is riskier on time 3 • StarCraft-High group > SC-Low at information filtering • Game factors such as success, and some offerings from RADAR to be taken into account • 2-months follow up Thank You! Todd Maddox Brad Love Marc Tomlinson Devon Greer Rusty Gomez Attention and StarCraft • Alerting • Sensitivity to incoming stimuli • Fronto-Parietal • Norepinephrine system, locus coeruleus • Orienting • Selection of information from sensory input • Cholinergic system, basal forebrain • Reduce neural activity and RT for invalid cues • Executive • Monitoring and resolving conflicts • Usual suspects: ACC and LatPFC • Targets of mesocortical dopamine system Training Novices Operating Span (OSpan) Ospan, Time3-Time1 O-Span Score 10 5 0 -5 -10 -15 -20 SC-Low SC-High SC Sims Task Switching Task Switching, Time2-Time1 Task Switching, Time2-Time1 0.10 0.08 0.04 RT Accuracy 0.06 0.02 0.00 -0.02 -0.04 -0.06 SC-Low SCHigh Switches SC 200 150 100 50 0 -50 -100 -150 -200 -250 -300 Sims SC-Low Non-Switches Switches SC Sims Non-Switches Task Switching, Time2-Time1 Task Switching, Time2-Time1 200 0.100 150 0.080 100 RT - Switch Cost ACC - Switch Cost SCHigh 0.060 0.040 0.020 0.000 50 0 -50 -100 -150 -0.020 -200 -0.040 -250 -0.060 -300 SCLow SCHigh SC Sims SC-Low SCHigh SC Sims Multitask Switching Multitask Switching, Time2-Time1 Multitask Switching, Time2-Time1 600.0 ACC - Cost RT - Cost 400.0 200.0 0.0 -200.0 -400.0 -600.0 SCLow SCHigh SC Sims 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 SC-Low SC-High SC Sims