Games 4 Health Conference, Boston
Game Accessibility Workshop
June 10th 2009
Game
Accessibility
Research @
Eelke Folmer

Overview
1.Introduction
2.Motivation for Research
3.Research Projects:
1 TextSL
4.Lessons Learned
5.Q&A

Eelke Folmer
Assistant Professor (2006)
Player-Game-Interaction Lab
2 MS. / 3 Ph.D / 1 PostDoc
Research Interests: Dead Space
 Human Computer Interaction
 Software Engineering
 Games / Virtual Worlds
 http://www.eelke.com Wii-Fit
 “You are what you play”:
BioShock

Motivations

Games are pervasive
Education
Advertising
Politics
Military Simulations
Health
...

Diverse audience
More & more people are interested in playing games
Varying amounts of experience & abilities
Individuals with disabilities often face significant
barriers to or are excluded from playing games
What are these barriers?
Illustrate with a Game Interaction Model

How do we play games?
game output 3 different
modalities
input
player

Detailed Steps
1.Game provides feedback {enemies/ gunfire/ sound}
2.Player decides what in-game response to provide:
1.(a) Do nothing.
2.(b) Move or reorient his/her character.
3.(c) Fire his/her gun.
4.(d) Change the weapon.
5.(e) Change the camera viewpoint.
6.(f) Reload, pause, or save the game.
3.The player must physically issue the chosen action(s)
through a physical device.
4.The internal state of the game may change and new
feedback may be provided

Synthesis
1. feedback 1. feedback 1. feedback
2. up/down 2. left/right/gas 2. left/right/shoot
3. button 3. wheel 3. button / mouse

Game Interaction model

Disability

Disability

Disability

Disability

Disability

Game Accessibility Stats
★Disability may affect ability to
play to different extents
★Based on Census 2006 data &
game interaction model we found
that in the US:
★6.2 Million disabled are
unable to play games at all
★25.9 Million disabled suffer
from reduced game experience
★Elderly overly represented in
our estimates.
★Nevertheless this will be a
big problem in the future

Our research
★Understand exactly what barriers
individuals with disabilities face.
★Research alternative ways to play:
★Feedback from alternative
modalities
★Reduced amounts of input.
★while retaining game experience
★Focus on game genres that are NOT
accessible
★More complex interaction models
★Provide examples & guidelines to
the game industry.

Guitar Hero
> 21M copies sold
Dominate sales charts
Not playable by

Viable strategies
?
VI accessible games Guitar Hero
• Speech • Interference with music
• Earcons
• Sonification • Too many stimuli
• Audio Cues

Haptic output
5 Different “notes”
!
Varying note lengths
Lookahead
<3 notes simultaneously
!
? ?
Glove
Pager engines (# ? )
“Haptic Cues”?
Position?

Tradeoffs
!
remove
lookahead
!
★“whack a mole” gameplay
Remove ★retain “guitar playing” feeling
1 button ★Easy to learn
★Fun!

Play testing

User Study
Participants
8 sighted + 4 visually impaired
age 20 - 41 (SD=4.5)
Groups
1 sighted Blind Hero first time players
2 blind Blind Hero first time players
3 sighted Blind Hero experts on Guitar Hero
4 sighted Guitar Hero first time players
Play 2 songs repeatedly (to measure
increase) for 4 times

Results
guitar Hero - novice
expert
VI
novice
Haptic is a viable strategy for Music based games
Performance lower but could increase over time
Most important: participants enjoyed the game

one switch games

Problem
switch controller
mouth controller
eye tracker one handed joystick
 rely upon adapted controllers
Constrained (type/ amount)
Suitable for arcade/puzzle games
Not suitable for popular game genres (FPS, RTS,
etc)

Research objective
less
?
Find smallest amount of input.
Preserve gameplay.
Find the “core” of the game.
What strategies can be used?
more

Experiences
less
✔
✔
✔
 Support all types of input
 Find minimum (may not be binary 1 switch)
 Identified the following mechanisms:
 automation
 reduction
more  scanning
 How do these change the gameplay?

Gordon’s Trigger Finger
Be able to respond fast
Scanning/automating/removing
Idea: player on top of a bot?
Evaluate alternatives
Scanning too slow moving user automated
Multiplayer only aiming user automated
fire user user
Railshooter
switch weapons user random
Balanced/ Fair open doors user automated
FUN jump/crawl user removed

Video

One Switch Interfaces
Music
Puzzle Games
virtual world
Sports Games first person shooter

TextSL

Virtual Worlds
Second Life / There / Active / Home
Higher degree of interaction.
Visually attractive
Entirely driven by user generated
content
Virtual Economy
Social Interaction most important
Estimated: in 4 years 80% of
internet users will have a VW
account

Accessibility Problem
?
? ?
Unlike Web, VW lack textual representation
Not Accessible with screen reader / tactile display
VW increasingly used as virtual classrooms /
distance learning (Section 508)
Communities for Disabled in SL (Virtual ability)
Research how VW can be made accessible
 develop prototype interface

Analyze differences
virtual worlds video games
 VW lack Combat/Storyline/Score
 No need for quick responses as you cannot die
 Everything is user generated (large amount of
content you cannot change)
 Viewer offers large number of functions including
creating objects. Many operations on objects. Objects
define their own functionality through scripts.
 VW viewers is more like a browser.

Output Strategies
 VI games (TerraFormers, AudioQuake, Shades of Doom) with
similar interaction mechanisms as VW.
 All use audio
 Levels have minimum number of objects
 Most strategies focus on being able to locate “enemies”
 Earcons & Audio Cues (too many objects in VW & you
cannot augment objects )
 Sound Radar (not applicable)
 Speech (applicable, but requires augmentation)

Controls
 Controls the same as the original games (arrow keys)
 Player has to do the pathfinding which is useful for mental
mapping (dungeons)
 Generic “use” command to interact with objects
 Design of levels accommodates limited interaction &
manual pathfinding
 CS difficult to implement in Virtual worlds as
 Large amount of objects & avatars
 Large number of functions to support.
 Mental mapping impossible in VW
 Manual Pathfinding is very hard

Different approach
 Command based interface (TextSL):
 Inspired by multi user dungeon games (MUD)
 “chat interface” to support social interaction
 Iteratively extract texts that can be read with a screen
reader
 Command based interaction supports interaction with large
amounts of objects & Avatars (“give my flower to jane”, “sit
on chair”)

TextSL features
 Commands in 3 categories:
 Exploration (move / fly / teleport / follow / describe /where)
 Communication (say / whisper / mute,.....)
 Interaction (sit, touch.....)
 Notable features:
 Natural language (go / travel / walk all map to “move” )
increasing learnability. Interpreter allows for prepositions and
adjectives “give my chair to...”
 Collision free navigation. e.g. “Move forward 100”
 Summarizer aims to minimize the amount of feedback
provided to avoid overwhelming the user with feedback.

User Studies
 Compare TextSL with SL
 8 Sighted & 8 Screen
reader users
 Tasks in 3 categories +
tutorial
 Measure success rate &
performance times
 TextSL has same succesrate as SL viewer for
performing tasks (accessibility)
 Exploration & Interaction significantly slower (usability)
 Communication the same.
 Command based interface is a feasible approach

Lessons Learned
 Research contributions:
 Haptic feedback is a viable strategy for making music
based games accessible to VI.
 Popular game genres can be made switch accessible
by automating, reducing or making it switch accessible
but it involves significant tradeoffs with regard to the
gameplay. For a number of popular game genres we
now understand these tradeoffs.
 Virtual worlds are different from Video games and
require a different strategy to make them accessible to
VI. A command-based interface is a feasible approach.
 lack of meta data for objects in SL

Questions?
A description of all research projects & media (video/
downloads) can be found on http://www.eelke.com