BT/PACCIT Event London, 13 Oct 2005

Download Report

Transcript BT/PACCIT Event London, 13 Oct 2005

Making Learning Fun
E-Drama Project
On-line dramatic improvisation (role-play) with
Artificial Intelligence enhancements
John Barnden, University of Birmingham
Amanda Oldroyd, BT Group CTO
Jemima Cattel, Hi8us Midlands Ltd
Jonnie Turpie, Maverick TV Ltd
Presentation Structure
Benefits of software
AI developments
User study feedback
3D interfaces
Introduction: What is E-drama
In this presentation, e-drama means dramatic improvisation or role-play where
the actors don’t work on a real stage but instead sit at computers and operate
virtual on-screen characters. Actors type in “speeches” which are shown in
speech bubbles. There is usually also a director who sits at another computer.
from a version
of the edrama
system from
Hi8us – already
a proven tool
Usually a
loose scenario
is specified,
such as a
or an
Science, etc.)
Seminar Themes Addressed
 Making learning more motivating and fun
 Challenging game-like activities
 Creativity
 Various types of hard-to-reach learner, and
shy or disaffected learners
 Next-generation e-learning resources
The Point of E-Drama
 Complements the use of live improvisation
in learning and creativity.
 It is anonymous and therefore less
intimidating, and also more motivating:
 92% of school students in recent test found
edrama less nervous-making
 81% found it more enjoyable overall
 Beneficial for emotionally difficult topics.
Plan of Rest of Talk
 Directorial interventions
 Desirability of semi-automating them
 Current developments towards this
 Points from recent user-testing
 More advanced graphics
Directorial Intervention
 Director can intervene to keep the drama on
fruitful tracks,
incl. with regard to expressed emotions.
 Director can send (written) messages to all
actors or “whisper” to an individual one.
 Director can introduce a bit-part character
such as a policeman to stir the action up.
A Reason for Directorial Intervention
 Emotional and other “affective” expression is
inappropriate, excessive or lacking.
-- Swearing or out-of-character rudeness
-- Excessive anger or hostility
-- Lack of expected hostility
-- Lack of expected embarrassment
-- Lack of expected concern about someone’s
emotional state
Artificial Intelligence and Human-Language
 Problem: demands on director are substantial.
 Partial automation would be beneficial.
 We are implementing
 automated detection of affect (e.g., emotion), hence
 automated directorial interventions, incl. introduction
of automated affect-sensitive bit-part characters
 Will allow more widespread and untrained use.
 Scalability, distributed interaction
Artificial Intelligence, contd
 The automation is not intended or
expected to make the overall
experience better.
 It would be a good result if it made it
not too much worse!
Automated Bit-Part Character
 EmEliza is our first research-prototype program
for operating an automated affect-sensitive bitpart character:
 It detects affective aspects of the other characters’
speeches, makes certain types of inference, and
makes short response speeches.
 Operating in our demo here.
EmEliza – A Bit-part Character Actor
EmEliza’s perception
of that character
I am scared of Mayid. He
is a bully.
I can't move my legs
and it's killing me! boo
hoo *cries*
I am going to hit you.
Why not let your mum
talk to his parents
Oh, dear, Please be
more brave. Help will be
here soon.
You look like a real
trouble maker.
User Testing: Pilot Run
Swanshurst School, Birmingham, May 2005
 Methodology:
39 subjects were involved
over two days.
Each of them participated in
both of the two selected
scenarios, namely
School Bullying and
Crohn’s Disease,
with/without EmEliza.
One Result:
Nobody suspected
that (sometimes) one
bit-part character was
Some Preliminary User-Testing Results
 When a certain character “Dave” was played by
 Little effect on the proportion of (other) actors who
regarded Dave as being poorly played.
 NB: substantial effect on various other measures and
on actors’ comments.
Having EmEliza as Dave, contd.
 Did raise the proportion of actors who got
bored from 9% to 18% …. BUT
 Reduced the proportion of actors saying
they disliked playing their own character
from 24% to 12%.
 Raised the proportion of actors who wanted
the improv to be longer from 74% to 88%.
Adding 3rd Dimension
 Dealing with sophisticated user groups:
 Experience with gaming technologies
 High expectations
 3d characters and customisation
 3d virtual environments
Realtime 3D Rendering
 Virtual world rendering
architecture (TARA):
realtime 3D engine for
avatars and backgrounds
 3D virtual characters with
customisation and
animation libraries
Expressive Behaviour - Demeanour System
 Generating expressive body language
for virtual characters in order to create
emotionally richer on-line experiences
 Behaviour is generated based on a
set of parameters representing
emotion, mood, personality etc.
 Customise characters’ behaviour in
Automated Behaviour linked to profiles
 Profiles related to the key
character types in the roleplay structure
 Each avatar animates
automatically in response to
other characters
 No user intervention
required during role-play
 Hi8us edrama is already a proven tool, used by a
range of users in a range of settings.
 Building on established role-play techniques – online.
 AI enhancements provide the opportunity to enable
more widespread and untrained use.
 3D enhancements build on existing software to
provide a richer user experience with low overhead.
Benefits of E-Drama:
from Learner Point of View
“It was really fun and a
change from ordinary
activities to help
understand bullying and
“I preferred using this to having a
conversation face to face because you can
almost ‘hide’ behind the character”
Benefits of E-Drama:
from Teacher Point of View
 Releases learner creativity, free
from peer pressure
 Encourages learner involvement
 Motivates study of the included
background material
 Low overhead for teachers to
become proficient in running
It’s fun for the teacher too!: can be guide, actor or director
Seminar Themes Addressed
 Making learning more motivating and fun
 Challenging game-like activities
 Creativity
 Various types of hard-to-reach learner, and
shy or disaffected learners
 Next-generation e-learning resources
Thank you