ESE 608 – Spring’04

“Knowledge Based Systems and Engineering Design” will focus on

INTELLIGENT AND ANIMATED SOFTWARE AGENTS:

A NEW TOOL FOR SYSTEMS SCIENCE

 

Prof. Barry G. Silverman

229c Towne Bldg, Tues. 4:30-7:30 PM

(215) 573-8368

barryg@seas.upenn.edu

 

This course will begin with an introduction to virtual reality personas and web-based agents, including their usage to assist, train, and entertain people wherever digital interfaces exist (on the web, in e-commerce, in games, in kitchen appliances, on your dashboard, etc.). What makes an agent rational? emotionally appealing? entertaining? We will explore mathematical theories of rationality and behavior, including those from cognitive, behavioral, and decision science. We will then progress into human behavior literature, personality and individual differences studies, and intelligent and emotive agent designs. We will examine various types of agents such as web shopping agents, emotive agents, personal support agents, chatterbots, mobile agents, virtual reality personas, game-based adversaries, pedagogical agent coaches, and multi-agent societies. Finally, students will learn principles about animation, simulated social interaction, speech generation, knowledge representation, agent planning and reasoning, agent communication languages, testing of the use of agent based systems, and methodologies/toolbenches for engineering of systems of intelligent and emotive agents.

 

Date

Topic

Assignment

Part I

1/13

1) Course overview

·          Overview this semester

·          Agents Defined, Taxonomy of Agents

·          Demo of PMFserv Environment

·          Graphics/Believability vs. Human Performance (PMFs)

 

1/20

2) Performance. Moderator Functions (PMFs) – Part I

·          Intro to PMF Anthology/Validity (& CMF mathematics)

·          Basic Theory: Inverted U Shape

·          Reservoirs/Virtual Stomach (calorie counting)

·          Depletion Factors (Exertion, Injury, etc.)

·          HANDS ON LAB (home assignment)

Silverman’03

 

Select PMF abstracts

 

 

1/27

3) Performance. Moderator Functions (PMFs) – Part II

·          Three types of Integrated Stress

·          Janis & Mann’s Conflict Theory (Coping Styles)

·          Impact of Stress on Perception & Decision Making

·          Intro to Affordance Theory (examine The Sims)

Turn in Lab 1

Gillis & Hursh

Janis & Mann

Silverman’03 (Affordnc Sect)

2/3

4) Emotion & Culture Modeling – Part I

·          Guidelines for Agent Social Interaction

·          Cognitive Appraisal theory & OCC Model

·          Modeling Culture via subjective utility

·          HANDS ON LAB (home assignment: cull literature for weights & markup affordances in OCC editor)

Silverman’01 (Emot. Sect.)

 

OCC Abstract

Damasio Abstr.

2/10

5) Emotion & Culture Modeling – Part II

·          Survey of Other OCC Implementations

·          Elliott’s Construal Frames

·          Ethnographic/Anthropologic Studies

Discuss Scenario Projects & Form Teams

Turn in Lab 2

 

Elliott’s Aff Reasnr

 

2/17

1st MIDTERM (Stress & Emotional Response of Agents)

Design Problem

Part II

2/24

7) Decision Layer of Agent Cognition

·          Axioms of Rationality vs. Bounded Rationality

·          Merging Stress and Emotion – Subjective Utility

·          Other Models of Decision Layer (RPD, etc.)

·          AI Models – Expert Systems, Planning Systems

Abstracts:

Prospect Theory

RPD-CBR

3/2

 

8) Combat Simulator Environments

·          Combat Game Environments (first person shooters vs. campaign games) – Editing character behaviors

·          Scenario Editing -- Unreal & HMS Videos

·          Demo of PMFserv Simulation World and Editors

·          Hands On Lab with PMFserv’s Editors

Proj.Plan Due

Silverman

3/9

BREAK WEEK

 

3/18

9) Agent Standards – Part I

·          Agent Oriented Programming (AOP) Overview

·          Mobile Agents & Mobile Agent Environments

·          Speech Act Theory

·          Agent Communication Languages (KQML, ACL, FIPA)

Lab 3 due

Bradshaw

3/23 GDC

NHAAP

10) Heart Sense Game and the AESOP Generator (R. Weaver)

·          Digital Casts & Role Playing Games

·          Narratology as a Game Mechanic

·          Edutainment & Heart Sense Game – Demo

·          AESOP Status Review

·          Hands On Lab with AESOP/HSG

AESOP

ICVS

3/30

 

11) Knowledge Representations & Standards Part II

·          State space, Problem Reduction, OO & Semantic Nets

·          Logic and Rules

·          Payoff Tables & Utility, Decision Trees & Influence

·          Ontologic Engineering/XML/RDF/DAML-OIL

 

4/6

12) Game Agents, Aesthetics, and Companion Agents

·          Design of Games and Game AI

·          Mechanics for Gameplay

·          Evaluation of Agents in Games & Measurement Issues

·          Playful Agents Embedded Elsewhere (Chatterbots, etc.)

Lab #4 (in class game design and reapir)

4/13

 

13) Cellular Automata, Agent Swarms, & Artificial Life (Jason Cornwell)

·        Theory of Evolution and Punctuated Equilibria

·        Artificial Life and Game of Life

·        Cellular Automata and Game Theory

·        ACASA-CA – Terrorism Event Spin

·        Consumer Buyer Behavior and Marketing Exposures

 lab 5 assigned

 

4/20

2nd MIDTERM

Evaluation topic

4/27

15) Reading Period

 

5/4

16) Scenario Project Presentations

Paper/PPT/Code

 

 

Readings

Silverman’03 – Tech Report on PMFserv, January 2003. Download from http://www.seas.upenn.edu/~barryg/PMFserv_2003_revisions.doc

 

Silverman’01 – Sect. 3.2 of Tech Report from Summer’01. Download from http://www.seas.upenn.edu/~barryg/TechRpt.pdf

 

PMF ‘Abstracts’ are usually in the various Course Documents Folders. If any are missing, they are to be Downloaded from the following file (open it and find the relevant abstract such as for physiological factors, Janis & Mann, OCC, Damasio, etc.) : http://www.seas.upenn.edu/~barryg/PMFset.zip

 

Other Readings In Course Document Folders:

Gillis & Hursh from CGF Conf. Proceedings

Elliott’s Affective Reasoner

Bradshaw et al., Agents for the Masses, IEEE Expert Mag

 

Scenario Project Description

Students in the course working in teams of about 2 or 3 each, will construct an intelligent animated agent (PMFserv environment to be provided by instructor). Each team should grapple with: scenario conceptualization, PMF parameter levels/stressors for the scenario, modeling the cultural values and emotive behavior of scenario participants, semantic markup of world (affordances), and what-if analysis of the results.  Thus there will be a given baseline scenario in the simulator for the PMFserv microworld. Together a team should construct one type of agent that will formulate beliefs about objects, situations, and other agents’ intentions, derive action plans, and interact intelligently with (against) the agents of other teams. Each agent will have a visible persona and inspectable reasoning processes.

 

Illustrative What-If Problems (Past Scenario Projects)

·        Skill level differences of agents (green, seasoned, burnt out)

·        Impact of chanting on crowd behavior

·        Impact of non-lethal weapons on agent PMFs

·        Individual differences in terms of stress tolerance

·        Cultural impacts upon behavior and personality

 

PLAN: Up to 5 pages (max) bounding the scenario what if questions, PMF factors to include, cultural value sets and emotive states to cull from the literature, affordance programming effort, and other tasks to complete the scenario study. Also the plan should include schedule, assigned tasks, and risks that need to be overcome.

 

FINAL PAPER/PRESENTATION: up to 25 page typed report (single space, 12 point, 1” margins), powerpoint, & demo. The result should describe the agent type created and tie it in to the literature reviewed.

 

 

Grading

 Midterm #1                        20%

 Midterm #2                        20%

 Class Participation              15%

 Homework/Labs              15%

 Scenario Project (plan, final paper, code, presentation)            30%

                                     

NOTE:

All homeworks (labs, cases, and plan) must be on time -- 10% per day loss in grade for each day late, no credit if 1 week late. Midterms and Scenario Project lose 1% of grade per day late.