Taking AI from the Laboratory to WalMart

1
Taking AI from the Laboratory to Wal-Mart

• What Game Players Think They See

2
What is BreakAway, Ltd.

• Develops advanced computer entertainment
  simulations (games) and migrates proven M,S,V
  tools to Gov/Military for Strategic/Operational
  training models and simulations for
• Experimentation Safe environment for failure
• Problem solving Quick what ifs
• Decision support Make vs. buy
• Analysis Law of unintended consequences
  interdependencies
• Training Resources/Cost containment
• Outcome Management MOEs

3
What I will Talk About

• How customers and players respond to AI
• What they expect from AI
• How you/developers/researchers can improve their
  response to AI
• Game design paradigms that might apply to
  Cognitive Science.

4
What I Base These Opinions On

• Responses to games by
• Focus Groups
• Beta Test Groups
• Customer Responses
• Personal Interactions with Customers
• Playing Games
• So all statements are just my opinions. There is
  absolutely no data behind any of these statements.

5
Public Transportation

• QWhy wont most people use public
  transportation?
• AMost people continue to use their car because
  it is the only time they have alone.

6
Game Design Rules

• One Decision Per Minute
• The Computer Shouldnt Have the Fun
• Decisions are only real if both options have a
  possible positive outcome in the players
  understanding of the decision.
• Users are better than the developers ALWAYS.

7
Symbolic Play

• Young players playing games they dont yet
  understand.
• Innovative strategies
• Unit Types put to other uses

8
Game Players vs AI

• Football AI vs. Randomness
• Players think the AI is smarter if it surprises
  them.
• Simply reacting properly isnt seen as
  intelligence
• A player may have a concept of intelligent play
  that doesnt match reality.
• Dont do something dumb repeatedly. It is ok to
  do dumb things occasionally.

9
Players want to believe

• Football Penalty Example
• Game players map their knowledge of the real
  world onto what they see on the screen
• Dr Mimi Chatterbot
• People get emotionally involved in the virtual
  world
• Path-finding Example
• Once players have formed a concept of what the
  computer is doing it is very hard to correct them.

10
Focus on What Players See Most

• Indy Racing Example
• Players are not professional drivers
• Players first impressions are important
• Football Animation
• Careful attention to what the players actually
  see is important

11
Dont Learn to be Dumb

• Football Coach AI learning
• Players arent acting realistically so learning
  from them is dubious.
• Be selective in what you learn. Players often
  have bizarre strategies.
• Realism of the simulation is directly related to
  the quality and type of learning that can be done.

12
Basic Points

• People notice intelligence when they are
  surprised by something.
• Trivial dumbness will overwhelm any actual
  smartness.
• Spend time on the things they see, not the things
  that are the most interesting challenges to you.
• Sometimes learning is bad.

13
Main Point

• Users want to believe a program is smart so
  dont do anything that will dissuade them
• Create and maintain the suspension of disbelief

14
Conclusions

• Doug Whatley dwhatley_at_breakawayltd.com
• Products should be designed around the consumer.