Download FPSP2004 - Computer Science & Engineering

Intelligent
Agents to
Deliver
Learning
Materials
Leen-Kiat Soh
Computer Science & Engineering
University of Nebraska
Lincoln, NE 68588-0115
[email protected]
http://www.cse.unl.edu/agents
Future Problem
Solving Workshop—
Hastings
January 21, 2004
Challenges
• Customized education
– e.g., Modularized courseware to meet specific
requirements, deficiencies, and sequences
• Adapting education
– e.g., Flexible courseware that adapts in real time
to student behavior and aptitude
• Effective distance education
– e.g., tools with anytime, anywhere capabilities
– e.g., infrastructures to bring classroom experience
to distance learners
Problem
• Corbett et al. (1999):
– “the arsenal of sophisticated computational modules
inherited from AI produce learning gains of approximately .3
to 1.0 standard deviation units compared with students
learning the same content in a classroom.”
• Graesser et al. (2001):
– “Human tutors produce impressive learning gains (between
.4 and 2.3 standard deviation units over classroom
teachers), even though the vast majority of tutors in a
school’s system have modest domain knowledge, have no
training in pedagogical techniques, and rarely use the
sophisticated tutoring strategies of ITSs.”
ITS = Intelligent Tutoring System
Problem 2
• Woolf et al. (2002) also lists abilities that are
needed or present in tutors:
– Generative:
• Generates appropriate instructional material (problems,
hints, help) based on student performance
– Student modeling:
• Assesses the current state of a student’s knowledge and
does something instructionally useful based on the
assessment
– Expert modeling:
• Models expert performance and does something
instructionally useful based on domain knowledge
Problem 3
• Woolf et al. (2002) also lists abilities that are
needed or present in tutors, cont’d:
– Instructional modeling:
• Changes pedagogical strategies based on the changing
state of the student model, prescriptions of an expert
model, or both
– Mixed-initiative
• Human Computer Interaction (HCI)
– Self-improving:
• Capacity to monitor, evaluate, and improve its own
teaching performance as a function of experience
Problem 4
• Graesser et al. (2001) criticize the current
state of tutoring systems:
– If students merely keep guessing until they find an action
that gets positive feedback, they can learn to do the right
thing for the wrong reasons – shallow learning
– The tutor does not ask students to explain their actions –
multiple choice questions
– The user interface of tutoring systems requires students to
display many of the details of their reasoning – no stepping
back to see the “basic approach”
– When students learn quantitative skills (e.g., algebra or
physics problem solving), they are usually not encouraged to
see their work from a qualitative, semantic perspective
Solution Ideas
• Examples
– PACT (Koedinger et al. 1997): algebra, geometry, and
computer languages
– ANDES (Gertner and VanLehn 2000; VanLehn 1996): physics
• Used as an adjunct to college and high-school physics courses to help
students do their homework problems
• Has an immediate feedback to enhance learning
– SHERLOCK (Lesgold et al. 1992): electronics
– ATLAS (VanLehn et al. 2000):
• Model-tracing
• Students scored significantly higher than the ANDES students on a
conceptual post-test
Solution Ideas 2
• Examples
– AutoTutor (Graesser et al. 2001): introductory
course in computer literacy
• Fundamentals of computer hardware, OS, Internet, etc.
• Was designed to be a good conversational partner that
comprehends, speaks, points, and displays emotions, all
in a coordinated fashion
• Simulates a multi-turn conversation to extract more
information from the student and get the student to
articulate pieces of the answer
• Pumps the student for what s/he knows before drilling
down to specific pieces of the answer
• Uses Latent Semantic Analysis (LSA) to compute
matches between the student’s speech acts to the
expectations
Solution Ideas 3
• Examples
– AutoTutor (Graesser et al. 2001): introductory
course in computer literacy, cont’d:
• Feeds back to the student at three levels:
• (a) backchannel feedback that acknowledges the
learner’s input
• (b) evaluative pedagogical feedback on the learner’s
previous turn based on the LSA values of the learner’s
speech acts (negative, neutral negative, neutral positive,
positive)
• (c) corrective feedback that repairs bugs and
misconceptions that learners articulate
Solution Ideas 4
• Examples
– CIRCISM (Freedman and Evens 1996)
– BEE (Basic Electricity and Electronics) tutor (Rosé
et al. 1999)
– EVELYN Reading Coach and EMILY Reading Coach
(Mostow and Aist 2002): Project LISTEN
• Help students read by listening to children read aloud
– BELVEDERE (Suthers et al. 2002):
• Supports students in collaboratively solving ill-structured
problems in science and other areas (such as public
policy) as they develop critical inquiry skills
Solution Ideas 5
• Teachable Agents (Biswas et al. 2002)
– let students teach the agents to do things;
through teaching, the students learn
• Articulate Software (Forbus 2002); properties
are:
– It should be fluent (some understanding of the
subject being taught)
– It should be supportive (scaffolding)
– It should be generative (pose new questions)
– It should be customizable (manually)
Solution Ideas: Intelligent Agents
• What is an agent?
– An agent is an entity that takes sensory input from
its environment, makes autonomous decisions,
and carries out actions that affect the environment
– A thermostat is an agent
– A calculator is not an agent
Agent
sensory
input
think!
Environment
output
actions
Solution Ideas: Intelligent Agents 2
• What is an intelligent agent?
– An intelligent agent is one that is capable of flexible
autonomous actions in order to meet its design
objectives, where flexibility means:
• Reactivity: agents are able to perceive their environment, and
respond in a timely fashion to changes that occur in order to
satisfy their design objectives
• Pro-activeness: agents are able to exhibit goal-directed
behavior by taking the initiative in order to satisfy their design
objectives
• Social ability: agents are capable of interacting with other
agents (and possibly humans) in order to satisfy their design
objectives
(Wooldridge and Jennings 1995)
Solution Ideas: Intelligent Agents 3
• Machine Learning in AI says
The acquisition of new knowledge and motor and cognitive skills
and the incorporation of the acquired knowledge and skills in
future system activities, provided that this acquisition and
incorporation is conducted by the system itself and leads to an
improvement in its performance.
• Agents that learn are intelligent
• Not all agents are intelligent!
Solution Ideas: Agent Environment
• Inaccessible vs. accessible
– Incomplete vs. complete data
• Deterministic vs. non-deterministic
– Certainty vs. uncertainty
• Episodic vs. non-episodic
– Each episode is independent or not
• Static vs. dynamic
– Remain unchanged except by the performance
of actions by the agent?
• Discrete vs. continuous
– “Chess game” vs. “taxi driving”
Solution Ideas: Why Agents?
• If the system-to-be-built has, during the
execution of the system
Why does a person hire an agent?
– Incomplete data
– Uncertainty in the assessment/interaction of its
environment
– Inter-dependent episodes of events
– No full control over the events in the environment
– An “open world”, instead of a “closed world”
• In other words, agents are used when you need
to build a system that is adaptive to an uncertain,
dynamic, and at times unexpected environment
– So you can make full use of the autonomous property
of an agent
Evaluation Criteria
• Pre-Design
Some examples, not exhaustive
– Necessity
• Is the agent solution necessary? Or an overkill?
– Feasibility
• Does the agent solution make sense? Is it impossible to
implement? Do we have the resources? Will it work?
• During Design
– Modularity
• Are there different modules and components? Is the “data”
separated from the “brain”?
– Extensibility
• What if we want to apply it to another problem?
– Scalability
• What if we want to apply it to the same but bigger problem?
Evaluation Criteria 2
• Post-Design
Some examples, not exhaustive
– Correctness
• Are the algorithms correctly designed? Are the
solutions correct?
– Usefulness
• Is the solution useful? Does it actually address the
problem? Does it help the user?
– Reliability
• Does the solution work for all possible problems?
Will the performance deteriorate after a while?
– Adaptiveness/Learning
• Can the solution evolve by itself to solve new
problems or to become better at solving old
problems?
AI
ILMDA
• Intelligent Learning Materials Delivery
Agent
Bibliography
•
•
•
•
•
•
Allen, J. F., D. K. Byron, M. Dzikovska, G. Ferguson, L. Galescu, and A. Stent
(2001). Toward Conversational Human-Computer Interaction, AI Magazine,
22(4):27-37.
André, E. and T. Rist (2001). Controlling the Behavior of Animated
Presentation Agents in the Interface: Scripting versus Instructing, AI
Magazine, 22(4):53-66
Birmingham, W. P., et al (1994. The University of Michigan Digital Library:
This Is Not Your Father's Library, Proc. Digital Libraries 94, Hypermedia
Research Laboratory, Texas A&M University, College Station, Tex., pp. 53-60.
Biswas, G., D. Schwartz, J. Bransford, and the Teachable Agents Group at
Vanderbilt (2002). Technology Support for Complex Problem Solving: From
SAD Environments to AI, in Forbus, K. D. and P. J. Feltovich (2002). (Eds.)
Smart Machines in Education, Menlo Park, CA: AAAI Press, pp. 71-97.
Cassell, J. (2001). Embodied Conversational Agents: Representation and
Intelligence in User Interfaces, AI Magazine, 22(4):67-83.
Cassell, J. and H. Vilhjálmsson (1999). Fully Embodied Conversational
Avatars: Making Communicative Behaviors Autonomous, Autonomous Agents
and Multi-Agent Systems, 2:45-64.
Bibliography 2
•
•
•
•
•
•
Cohen, P. R. and S. L. Oviatt (1995). The Role of Voice Input for HumanMachine Communication, in Proceedings of the National Academy of
Sciences, 92(22):9921-9927.
Corbett, A., J. Anderson, A. Graesser, K. Koedinger, and K. VanLehn (1999).
Third Generation Computer Tutors: Learn from or Ignore Human Tutors? in
Proceedings of the 1999 Conference of Computer-Human Interaction, 8586.
D’Souza, A., J. Rickel, B. Herreros, and W. L. Johnson (2001). An
Automated Lab Instructor for Simulated Science Experiments, in J. D.
Moore, C. L. Radfield, and W. L. Johnson (eds.) Artificial Intelligence in
Education: AI-ED in the Wired and WirelessFuture, 65-76, Amsterdam: IOS.
Forbus, K. D. (2002). Articulate Software for Science and Engineering
Education, in Forbus, K. D. and P. J. Feltovich (2002). (Eds.) Smart
Machines in Education, Menlo Park, CA: AAAI Press, pp. 235-267.
Forbus, K. D. and P. J. Feltovich (2002). (Eds.) Smart Machines in
Education, Menlo Park, CA: AAAI Press.
Freedman, R. and M. W. Evens (1996).
Generating and Revising
Hierarchical MultiTurn Text Plans in an ITS, in Intelligent Tutoring Systems:
Proceedings of the 1996 Conference, 632-640.
Bibliography 3
•
•
•
•
•
Gertner, A. S. and K. VanLehn (2000). ANDES: A Coached Problem-Solving
Environment for Physics, in Intelligent Tutoring Systems: Fifth International
Conference, ITS 2000, eds. G. Gautheier, C. Frasson, and K. VanLehn, 133142.
Graesser, A. C., K. Wiemer-Hastings, P. Wiemer-Hastings, R. Kreuz, and the
Tutoring Research Group (1999). AutoTutor: A Simulation of a Human Tutor,
Journal of Cognitive Systems Research, 1(1):35-51.
Graesser, A. C., K. VanLehn, C. P. Rosé, P. W. Jordan, and D. Harter (2001).
Intelligent Tutoring Systems with Conversational Dialogue, AI Magazine,
22(4):39-51.
Herzog, M. and P. Petta (1995). Multimedia Information System in OpenWorld Domains, in Working Notes of 1st Int. Workshop on Multimedia
Information Systems, Arlington, VA, September 28-30, 13-16.
Herzog, M. and P. Petta (1996). Knowledge Representation in WERKL, an
Architecture for Intelligent Multimedia Information Systems, in Workshop
Notes of the ECAI'96 Workshop on Knowledge Representation for Interactive
Multimedia Systems: Research and Experience (KRIMS'96), ECAI'96,
•
Budapest, Hungary, August 12, 17-25.
Johnson, W. L. (2001). Pedagogical Agent Research at CARTE, AI Magazine,
22(4):85-94.
Bibliography 4
•
•
•
•
•
•
Johnson, W. L., J. W. Rickel, and J. C. Lester (2000). Animated Pedagogical
Agents: Face-to-Face Interaction in Interactive Learning Environments,
International Journal of Artificial Intelligence in Education, 11:47-78.
Koedinger, K. R., J. R. Anderson, W. H. Hadley, and M. A. Mark (1997).
Intelligent Tutoring Goes to School in the Big City, Journal of the Artificial
Intelligence in Education, 8(1):30-43.
Lesgold, A., S. Lajoie, M. Bunzo, and G. Eggan (1992). SHERLOCK: A
Coached Practice Environment for an Electronics Troubleshooting Job, in J. H.
Larkin and R. W. Chabay (eds.) Computer Assisted Instruction and Intelligent
Tutoring Systems, 201-238, Hillsdale, NJ: Lawrence Erlbaum.
Lester, J. C., C. B. Callaway, J. P. Grégoire, G. D. Stelling, S. G. Towns, and L.
S. Zettlemoyer (2002). Animated Pedagogical Agents in Knowledge-Based
Learning Environments, in Forbus, K. D. and P. J. Feltovich (2002). (Eds.)
Smart Machines in Education, Menlo Park, CA: AAAI Press, pp. 269-298.
Marsella, S. and W. L. Johnson (1998). An Intelligent Assistant for Team
Training in Dynamic Multi-Agent Virtual Worlds, in Proceedings of Intelligent
Tutoring Systems, 4th International Conference, 464-473.
Mostow, J. and G. Aist (2002). Evaluating Tutors that Listen: An Overview of
Project LISTEN, in Forbus, K. D. and P. J. Feltovich (2002). (Eds.) Smart
Machines in Education, Menlo Park, CA: AAAI Press, pp. 169-234.
Bibliography 5
•
•
•
•
•
•
Raines, T., M. Tambe, and S. Marsella (2000). Automated Assistants to Aid
Humans in Understanding Team Behaviors, in Proceedings of the Fourth
International Conference on Autonomous Agents, 419-426.
Rickel, J. and W. L. Johnson (1999). Animated Agents for Procedural
Training in Virtual Reality: Perception, Cognition, and Motor Control, Applied
Artificial Intelligence, 13(4):343-382.
Rickel, J., R. Ganeshan, C. Rich, C. L. Sidner, and N. Lesh (2000). TaskOriented Tutorial Dialogue: Issues and Agents, paper presented at the AAAI
Fall Symposium on Building Dialogue Systems for Tutorial Applications,
November 3-5, North Falmouth, MA.
Rosé, C. P., B. Di Eugenio, and J. Moore (1999). A Dialogue-Based Tutoring
System for Basic Electricity and Electronics, in S. P. Lajoie and M. Vivet (eds.)
Artificial Intelligence in Education, 759-761, Amsterdam: IOS.
Rusell, S. and P. Norvig (1995). Artificial Intelligence: A Modern Approach,
Prentice-Hall.
Schank, R. and A. Neaman (2002). Motivation and Failure in Educational
Simulation Design, in Forbus, K. D. and P. J. Feltovich (2002). (Eds.) Smart
Machines in Education, Menlo Park, CA: AAAI Press, pp. 37-69.
Bibliography 7
•
•
•
•
•
•
•
Shaw, E., W. L. Johnson, and R. Ganeshan (1999). Pedagogical Agents on
the Web, in Proceedings of the Third International Conference on
Autonomous Agents, 283-290.
Smith, J. (2000). GrandChair: Conversational Collection of Family Stories, M.
S. Thesis, Media Lab, Massachusetts Institute of Technology, Cambridge, MA.
VanLehn, K. (1996). Conceptual and Metalearning during Coached Problem
Solving, in Proceedings of the Third Intelligent Tutoring Systems Conference,
29-47, Berlin: Springer-Verlag.
Suthers, D., J. Connelly, A. Lesgold, M. Paolucci, E. E Toth, J. Toth, and A.
Weiner (2002). Representational and Advisory Guidance for Students
Learning Scientific Inquiry, in Forbus, K. D. and P. J. Feltovich (2002). (Eds.)
Smart Machines in Education, Menlo Park, CA: AAAI Press, pp. 7-35.
Weiss, G. (ed.) (1999). Multiagent Systems: A Modern Approach to
Distributed Artificial Intelligence, MIT Press.
Wooldridge, M. and N. R. Jennings (1995). Intelligent Agents: Theory and
Practice, The Knowledge Engineering Review, 10(2):115-152.
Woolf, B. P., J. Beck, C.Eliot, and M. Stern (2002). Growth and Maturity of
Intelligent Tutoring Systems: A Status Report, in Forbus, K. D. and P. J.
Feltovich (2002). (Eds.) Smart Machines in Education, Menlo Park, CA: AAAI
Press, pp. 99-144.
Latent Semantic Analysis
• Latent Semantic Analysis (LSA) is a theory and method for extracting
and representing the contextual-usage meaning of words by
statistical computations applied to a large corpus of text (Landauer
and Dumais, 1997). The underlying idea is that the aggregate of all
the word contexts in which a given word does and does not appear
provides a set of mutual constraints that largely determines the
similarity of meaning of words and sets of words to each other.
– Based on Landauer, T. K., P. W., and D. Laham (1998) Introduction to
Latent Semantic Analysis, Discourse Processes, 25:259-284.