Вы находитесь на странице: 1из 23

An Approach to a Knowledge Reconstruction Engine

for Supporting Event Planning

Shigeki AMITANI, Mikihiko MORI and Koichi HORI


Department of Advanced Interdisciplinary Studies,
The University of Tokyo

2003/4/16 AAAI2002 1
Outline

• Background & Problems


• Our Approach
• Experiments & Methodology
• Examples of Knowledge Reconstruction
• System Image & Expected Interaction
• Future Work

2003/4/16 AAAI2002 2
Background

• Event Planning:
– Designing process with implicit knowledge of an experienced
planner
• Evaluation of an Event’s Effects:
– Questionnaires (5-point scale, free-answer)
• Feedback to planners:
– Statistical data and visitors’ comments

2003/4/16 AAAI2002 3
Problems

• Planners cannot obtain strategic knowledge:


– Statistical data do not tell why the results were obtained.
= lacking the contexts
• There is no systematic feedback to planners:
– It is difficult to accumulate and utilize gathered information as
"knowledge" on event-plannings.
• No accountability to clients:
– Because no clear strategy except for planners' tacit knowledge

2003/4/16 AAAI2002 4
To Utilize Knowledge: "Knowledge Reconstruction"
"Knowledge is embedded in a context of interaction
between human beings and artifacts"
It is impossible to capture or accumulate It is necessary to know
"knowledge" itself, but it is possible to “a context where
observe and obtain contexts in which knowledge is produced”
knowledge is produced and which can be
to utilize knowledge
a trigger to create new knowledge.
[Nakakoji, Hori, 2002] [Fischer, 2001]
Create New Knowledge

Obtain information Utilize Knowledge

with its context


Knowledge Reconstruction
2003/4/16 AAAI2002 5
Our Approach

To support event planning process...


• Methodology for Articulation of:
– planners’ intentions
– visitors’ impressions
– gaps between them to discover what causes what type of
effect

• Supporting System for:


– understanding "contexts" generated at real event sites
– discovering something interesting and useful for future
plannings
– promoting both planners' and clients' understanding

2003/4/16 AAAI2002 6
Experiments

• World PC Expo 2001 (19-22 Sep. ’01)


– FUJI Xerox
– Canon
– Toshiba
• Tokyo Motor Show 2001 (26 Oct. – 7 Nov. ’01)
– SUBARU
(both in Makuhari Messe, Japan)

2003/4/16 AAAI2002 7
Methodology

• Planners’ Intention:
à “What did you implement to express event concepts?”
à “How did you implement event objects to express the concepts?”
• Visitors’ Impression:
– Let subjects browse in a booth / booths with a recording device
– Interview with “Retrospective Report Method” with a visual aid
à “What did you look at?”
“What did you think about it?”
“Then how did you behave?”

à Compare them
2003/4/16 AAAI2002 8
A Recording Unit and Protocol Data

DV Camera

developed (?) at AI Lab, RCAST

time Percieved Object Thought Action


17.45 Explanation of the booth → I don't care what it is
18.16 I'm filling in a questionnaire if I can get something
→ Fill the questionnaire
19.38 I finished writing → Stand up
19.50 I'm going to look at the explained commodity → Move to an exhibision corner
20.30 I want to ask how much it is → Look at a staff

2003/4/16 AAAI2002 9
An Example Protocol

• At Motor Show: Interaction with the other visitors

– “A companion took a picture with a family. Both of


the companion and the child smiled. My (= the
subject's) children also like cars. They would be
delighted if I took them here. That is a good idea.”

à This comment caused a new strategy

2003/4/16 AAAI2002 10
A New Strategy was Produced

• A planner hit upon a new strategy:

– “By inviting customers’ families, the other visitors can


feel in the way mentioned. Moreover, the invited family
will also feel better because they feel “they are invited
as special guests” and this family can enjoy being a
customer of the company, which will be great benefit to
the company, too.”

2003/4/16 AAAI2002 11
A Good Example of Knowledge Reconstruction

• Following pieces of knowledge are not anything new, but


their combination results in a new strategy.
– There are interactions among visitors.
– People will be glad if they are invited as “special guests”.
– If visitors like the company, it is benefit for the company.
– And so on

àTo create and utilize information as “knowledge”, it should be


accompanied with “contexts where knowledge was produced”
which normal statistical data lack.

2003/4/16 AAAI2002 12
Another Example

• Following phenomenon was observed:


– A subject was impressed by a speedy leaser printer at FUJI
Xerox booth.
– Next he went to Canon booth with looking for leaser printers,
though no leaser printer was exhibited at Canon booth.
– Though he was looking for a leaser printer all the way, he could
not find one (of course).
– Then he was disappointed with Canon booth and left the booth.

2003/4/16 AAAI2002 13
New Type of Knowledge was Obtained

• According to planners, this phenomenon was new and


impressive to them because they do not think about
what are exhibited at neighbor booths.

à This phenomenon was caused by “a visitor’s context”. Our


approach successfully revealed what was unobtainable with
traditional investigation method.

2003/4/16 AAAI2002 14
System Requirements

• To discover something interesting from a large amount of


data
• To construct a case base of findings for future utilization
• To promote planners reflective thinking to create new
knowledge
• To explain in a persuasive way to promote clients’
understanding

2003/4/16 AAAI2002 15
System Components

• ChronoSpace
– A tool for browsing a visitor’s protocol in a microscopic way
• ContextMap
– A tool for browsing “interesting phenomena” in a
macroscopic way and creating new knowledge

In creative activities, it is necessary to move between overview and detail.

2003/4/16 AAAI2002 16
ContextMap
System Image All focused data are arranged

Event object name

ChronoSpace Comment Area


Objects and visitor’s flow
Protocol Area

: A point where an event


object is located
: A point where a verbal
report was produced on
a located object
: A point where a verbal
report was produced
Locations of objects are calculated along with
their similarity in terms of their effect on the
visitors’ mental impressions
2003/4/16 AAAI2002 17
Expected Interactions(1/3)

Reflective Thinking

– Using ChronoSpace, browse each data after the investigation to


grasp what / how visitors really looked at / thought / acted
– Find something interesting / unexpected either in a positive way or
negative way and put tags on nodes
– Accumulate remarkable phenomena

2003/4/16 AAAI2002 18
Expected Interactions(2/3)

Supporting Planning Phase

– Case study: A user can browse what caused interesting phenomena


on ContextMap
– Modify a two-dimensional space to fit object positions to the user’s
mental space
– Abstract a concept by grouping and labeling objects

2003/4/16 AAAI2002 19
Expected Interactions(3/3)

Persuading Clients

– Explanation by indicating all of notable past instances

2003/4/16 AAAI2002 20
Future Work

• Develop the system prototype


• Have a planner use the system prototype and observe
his/her usage
• Interaction Design for real workflow (especially designing
interaction of ContextMap)
• Refine the system

2003/4/16 AAAI2002 21
acknowledgement

we really appreciate for kind help of:

… ms. hiroko shoji and mr. hirohito shibata


… members of our laboratory
...mr. brian clarkson @ mit media lab.
... ms. ueoka @ icsl, rcast
… Dentsu Inc.
… and all of the people who gave us supports

2003/4/16 AAAI2002 22
Questions?

• Slowly
• Clearly
• Loudly
…please!

2003/4/16 AAAI2002 23

Вам также может понравиться