Exploring the Constraints of Human Behavior Representation

blabbedharborΤεχνίτη Νοημοσύνη και Ρομποτική

23 Φεβ 2014 (πριν από 3 χρόνια και 3 μήνες)

60 εμφανίσεις

April 28, 2004

John C. Giordano


Masters Project Presentation

1

Exploring the Constraints of
Human Behavior Representation

A Masters Project Presentation

John C. Giordano

Prof. Paul Reynolds
-

Advisor

April 28, 2004

John C. Giordano


Masters Project Presentation

2

Outline


Introduction & problem statement


Key terms


Highlights of literature review


Findings


Proposed framework for considering
human behavior representation (HBR)
capabilities


Conclusions

April 28, 2004

John C. Giordano


Masters Project Presentation

3



“We can only see a short distance ahead, but we
can see plenty there that needs to be done.”



Alan Turing


Computing Machinery and Intelligence


In
Mind:
236, 1950

April 28, 2004

John C. Giordano


Masters Project Presentation

4

Introduction


In 1950, Alan Turing proposes the
Imitation
Game



Machines competing with or replacing humans



Human behavior representation (HBR) refers to
the
portrayal

of humans



HBR is not Artificial Intelligence


More constrained


Still a challenge


April 28, 2004

John C. Giordano


Masters Project Presentation

5

What is the Problem?


HBR is critical to many, but has proven elusive



Several large
-
scale development failures with
prominent HBR requirements


DoD’s Joint Simulation System (JSIMS)


NASA’s Air Traffic Management (ATM) simulation



Shortcomings noted by many in the community



April 28, 2004

John C. Giordano


Masters Project Presentation

6

How We Attempt to Address It


Examined successes and failures in research,
design and implementation



Describe what is currently attainable and
propose what is unachievable



Present a framework for assessing HBR
capabilities



Seeking publication of research conducted to
date

April 28, 2004

John C. Giordano


Masters Project Presentation

7

Key Terms


HBR:

a computer
-
based model that mimics either the
behavior of a single human or the collective action of a
team of humans



Intelligent Software Agent:

an artificial agent that
operates in a software environment and imitates human
intelligence by mechanical means in pursuit of the goals
of its clients



Human Cognition:

the process of receiving, processing,
storing, and using information in humans

April 28, 2004

John C. Giordano


Masters Project Presentation

8

Literature Review


Over 60 publications (papers, journal
articles, texts, tech reports, requirements
documents)



Extended annotated bibliography



Thorough, but not fully exhaustive

April 28, 2004

John C. Giordano


Masters Project Presentation

9

Literature Review


Modeling Human and Organizational Behavior.

Richard
W. Pew and Anne S. Mavor (eds.). National Academy
Press, Washington DC, 1998.



Techniques for Modeling Human Performance in
Synthetic Environments: A Supplemental Review.

Frank E. Ritter, et al. Human Systems Information
Analysis Center, Wright Patterson AFB, OH, 2002.



A Taxonomy of Human Behavior Representation
Requirements.

Scott Y. Harmon. 11
th

Conference on
Computer Generated Forces and Behavior
Representation, 2002.

April 28, 2004

John C. Giordano


Masters Project Presentation

10

Harmon’s Taxonomy

Human

Representation

Non
-
Cognitive

Factors

Cognitive

Capabilities

Application

Functions

April 28, 2004

John C. Giordano


Masters Project Presentation

11

Harmon’s Taxonomy

April 28, 2004

John C. Giordano


Masters Project Presentation

12

Harmon’s Taxonomy

April 28, 2004

John C. Giordano


Masters Project Presentation

13

Findings

April 28, 2004

John C. Giordano


Masters Project Presentation

14

Findings


The tools used to model and simulate HBR are
constrained



The phenomena associated with HBR are highly
complex



At times, HBR requirements vastly exceed capabilities



Capabilities and constraints should be clearly articulated
to the community



Some capabilities may (only) be attained with the
emergence of a disruptive technology

April 28, 2004

John C. Giordano


Masters Project Presentation

15

Some Tools




Soar: general cognitive architecture for
intelligent agents



COGNET/iGEN: emulator for human
decision
-
making and problem
-
solving



ACT
-
R: architecture for human cognition

April 28, 2004

John C. Giordano


Masters Project Presentation

16

Three Categories of HBR


Capabilities

Mature

Developing

Unachievable in
practice

Constrained

speech

recognition,

parsing

and

generation

X

Course

of

Action

(COA)

analysis,

selection

and

implementation

X

Rudimentary

emotions

X

Human

physiological

characteristics

X

Semi
-
automated

coarse
-
grained

behavior

generation

X

Probabilistic

human

performance

simulation

and

prediction

X

Autonomous,

convincing

group

behavior

X

COA

generation

X

Interdependence

between

physiology,

emotion

and

cognition

X

Behavior

adaptation

appropriate

to

dynamic

scenarios

X

X

Speech

generation

w/

appropriate

prosody

X

X

Pattern

recognition

coupled

w/

appropriate

decision
-
making

X

X

Generalized

behavior

prediction

X

X

A

single

framework

for

modeling

human

behavior

at

multiple

levels

of

resolution

X

X

Complex

cognition,

reasoning

and

learning

X

Conversational

dialogue

X

Synthesis

of

autonomous

knowledge

acquisition,

planning

and

behavior

X

Complete

integration

between

emotion,

cognition

and

behavior

X

April 28, 2004

John C. Giordano


Masters Project Presentation

17



Constrained speech recognition, parsing and generation


Course of Action (COA) analysis, selection and
implementation


Rudimentary emotions


Human physiological characteristics


Semi
-
automated coarse
-
grained behavior generation


Probabilistic human performance simulation and
prediction

Mature Capabilities

April 28, 2004

John C. Giordano


Masters Project Presentation

18



Autonomous, convincing group behavior




COA generation




Interdependence between physiology,
emotion and cognition

Developing Capabilities

April 28, 2004

John C. Giordano


Masters Project Presentation

19



Behavior adaptation appropriate to dynamic
scenarios



Speech generation w/ appropriate prosody



Pattern recognition coupled w/ appropriate
decision
-
making



Generalized behavior prediction



A single framework for modeling human
behavior at multiple levels of resolution

Developing and Unachievable

April 28, 2004

John C. Giordano


Masters Project Presentation

20


Complex cognition, reasoning and learning


Conversational dialogue


Synthesis of autonomous knowledge
acquisition, planning and behavior


Complete integration between emotion,
cognition and behavior

Unachievable in Practice

April 28, 2004

John C. Giordano


Masters Project Presentation

21

A Generational Framework for

Considering HBR Capabilities


Analogous to the generations of programming
languages



Generations


1
st
:

speech recognition, rudimentary emotions/physiology,
probabilistic performance


2
nd
:

domain
-
independent speech, COA generation, adaptive
behaviors


3
rd
:

single cognitive framework, architecture for multi
-
resolution
behavior modeling, etc.


4
th
:

approaching human faculties

April 28, 2004

John C. Giordano


Masters Project Presentation

22

A Generational Framework for

Considering HBR Capabilities

General

Modeled Phenomenon


Measurable


Not Measurable

Model Specificity

Speech recognition

Integrated emotion and cognition

Conversational dialogue

Autonomous learning and planning

Concrete

April 28, 2004

John C. Giordano


Masters Project Presentation

23

Conclusions



Turing foresaw human
-
machine
competition


HBR comprises portrayal



Requirements development needs
improvement



Opportunities for continued research

April 28, 2004

John C. Giordano


Masters Project Presentation

24

Future Work


Update and extend Pew and Mavor, Ritter
et al.



Focus on HBR successes, particularly with
promise of generalization



Rigorous, formalized HBR requirements

April 28, 2004

John C. Giordano


Masters Project Presentation

25

Questions?


“It takes some philosophical discipline, in
short, to resist specious blurrings of
differences between simulations and the
phenomena they simulate.”



Larry Crockett



In
The Turing Test and the Frame Problem: AI’s Mistaken Understanding of Intelligence



Intellect Books, 1994