Noboru Matsuda Human-Computer Interaction Institute

Slides:



Advertisements
Similar presentations
Introduction to AuthorIT April 10, 2006 Symposium on Knowledge Representation TICL SIG Joseph M. Scandura, Ph.D. Chairman, Board Scientific Advisors, MERGE.
Advertisements

Modelling with expert systems. Expert systems Modelling with expert systems Coaching modelling with expert systems Advantages and limitations of modelling.
Principal Component Analysis Based on L1-Norm Maximization Nojun Kwak IEEE Transactions on Pattern Analysis and Machine Intelligence, 2008.
What can CTAT do for you? Overview of the CTAT track Vincent Aleven, Bruce McLaren and the CTAT team 3rd Annual PSLC LearnLab Summer School Pittsburgh,
Educational data mining overview & Introduction to Exploratory Data Analysis Ken Koedinger CMU Director of PSLC Professor of Human-Computer Interaction.
Improving learning by improving the cognitive model: A data- driven approach Cen, H., Koedinger, K., Junker, B. Learning Factors Analysis - A General Method.
An Individualized Web-Based Algebra Tutor D.Sklavakis & I. Refanidis 1 An Individualized Web-Based Algebra Tutor Based on Dynamic Deep Model Tracing Dimitrios.
Supporting (aspects of) self- directed learning with Cognitive Tutors Ken Koedinger CMU Director of Pittsburgh Science of Learning Center Human-Computer.
Does Conjunctive Knowledge Tracing Provide Leverage to the Temporal and Location Heuristics in Error Attribution? Adaeze Nwaigwe University of Maryland.
Learning from Observations Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 18 Spring 2004.
A C OMPUTATIONAL M ODEL OF A CCELERATED F UTURE L EARNING THROUGH F EATURE R ECOGNITION Nan Li Computer Science Department Carnegie Mellon University Building.
Learning from Observations Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 18 Fall 2005.
LEARNING FROM OBSERVATIONS Yılmaz KILIÇASLAN. Definition Learning takes place as the agent observes its interactions with the world and its own decision-making.
Learning from Observations Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 18 Fall 2004.
SAT-Based Decision Procedures for Subsets of First-Order Logic
6.2 Modelling the Learner ISE554 The WWW for eLearning.
Our Explorations with CTAT!
LEARNING FROM OBSERVATIONS Yılmaz KILIÇASLAN. Definition Learning takes place as the agent observes its interactions with the world and its own decision-making.
The Implicit Mapping into Feature Space. In order to learn non-linear relations with a linear machine, we need to select a set of non- linear features.
+ Doing More with Less : Student Modeling and Performance Prediction with Reduced Content Models Yun Huang, University of Pittsburgh Yanbo Xu, Carnegie.
Educational Data Mining Overview John Stamper PSLC Summer School /25/2011 1PSLC Summer School 2011.
Cognitive Science Overview Design Activity Cognitive Apprenticeship Theory Cognitive Flexibility Theory.
Learner Self-Correction in Solving Two-Step Algebraic Equations Brandy C. Judkins, School of Professional Studies in Education, Johns Hopkins University.
DataShop: An Educational Data Mining Platform for the Learning Science Community John Stamper Pittsburgh Science of Learning Center Human-Computer Interaction.
How to solve using cross multiplication Created by: Brittany and Andrea.
Intelligent Tutoring Systems Traditional CAI Fully specified presentation text Canned questions and associated answers Lack the ability to adapt to students.
31 st October, 2012 CSE-435 Tashwin Kaur Khurana.
Beverly Park Woolf University of Massachusetts/Amherst U.S.A
Concept Attainment Inquiry Lessons.  Is used to teach concepts, patterns and abstractions  Brings together the ideas of inquiry, discovery and problem-solving.
Improving a Mathematical Intelligent Tutoring System Experiments & Equation Solver Improvements July 27th, 2012 Jennifer Ferris-Glick & Hee Seung Lee.
Made with Protégé: An Intelligent Medical Training System Olga Medvedeva, Eugene Tseytlin, and Rebecca Crowley Center for Pathology Informatics, University.
Introduction to the Cognitive Tutor Authoring Tools (CTAT) and Example-Tracing Tutors Bruce McLaren Systems Scientist, Co-Manager of the CTAT Project Team.
Author: James Allen, Nathanael Chambers, etc. By: Rex, Linger, Xiaoyi Nov. 23, 2009.
Teaching a Practical Skill
 Knowledge Acquisition  Machine Learning. The transfer and transformation of potential problem solving expertise from some knowledge source to a program.
Prompts to Self-Explain Why examples are (in-)correct Focus on Procedures 58% of explanations were procedure- based Self-explanation is thought to facilitate.
Sampletalk Technology Presentation Andrew Gleibman
Simulated Student: Building Cognitive Model by Demonstration Noboru Matsuda School of Computer Science Carnegie Mellon University.
Measuring What Matters: Technology & the Assessment of all Students Jim Pellegrino.
1 Multimedia-Supported Metaphors for Meaning Making in Mathematics Moreno & Mayer (1999)
Intelligent Database Systems Lab N.Y.U.S.T. I. M. A Coursework Support System for Offering Challenges and Assistance by Analyzing Students’ Web Portfolios.
Spring 2011 Tutor Training Modern Learning Theories and Tutoring Designed and Presented by Tem Fuller.
DCA Consortium Group Session 2 Erie 1 BOCES. Agenda  State Updates from Network Team Institutes  Testing information  Upgrading Sample Tasks  Goals.
SimStudent: A computational model of learning for Intelligent Authoring and beyond Noboru Matsuda Human-Computer Interaction Institute Carnegie Mellon.
SD modeling process One drawback of using a computer to simulate systems is that the computer will always do exactly what you tell it to do. (Garbage in.
Prerequisite Skills VOCABULARY CHECK 40 ANSWER What is the least common denominator of and ? Which equation is a direct variation equation,
Applying the Redundancy Principle ( Chapter 7) And using e-learning data for CTA Ken Koedinger 1.
SimStudent: Building a Cognitive Tutor by Teaching a Simulated Student Noboru Matsuda Human-Computer Interaction Institute Carnegie Mellon University.
Data mining with DataShop Ken Koedinger CMU Director of PSLC Professor of Human-Computer Interaction & Psychology Carnegie Mellon University.
1 Learning through Interactive Behavior Specifications Tolga Konik CSLI, Stanford University Douglas Pearson Three Penny Software John Laird University.
Mass Producing Example- Tracing Tutors Bruce McLaren Human-Computer Interaction Institute Carnegie Mellon University.
System To Generate Test Data: The Analysis Program Syed Nabeel.
Of An Expert System.  Introduction  What is AI?  Intelligent in Human & Machine? What is Expert System? How are Expert System used? Elements of ES.
George Goguadze, Eric Andrès Universität des Saarlandes Johan Jeuring, Bastiaan Heeren Open Universiteit Nederland Generation of Interactive Exercises.
A M ACHINE L EARNING A PPROACH FOR A UTOMATIC S TUDENT M ODEL D ISCOVERY Nan Li, Noboru Matsuda, William Cohen, and Kenneth Koedinger Computer Science.
Finding Answers. Steps of Sci Method 1.Purpose 2.Hypothesis 3.Experiment 4.Results 5.Conclusion.
July 8, 2008In vivo experimentation: 1 Step by Step In Vivo Experimentation Lecture 3 for the IV track of the 2011 PSLC Summer School Philip Pavlik Jr.
Solving 2 step equations. Two step equations have addition or subtraction and multiply or divide 3x + 1 = 10 3x + 1 = 10 4y + 2 = 10 4y + 2 = 10 2b +
Learning Analytics isn’t new Ways in which we might build on the long history of adaptive learning systems within contemporary online learning design Professor.
 Knowledge Acquisition  Machine Learning. The transfer and transformation of potential problem solving expertise from some knowledge source to a program.
Creating Physicists: Making reasoning explicit: metacognition and the relative value of evidence
Direct Instruction Model
Inquiry learning and SimQuest
Solve for variable 3x = 6 7x = -21
Solving Systems Check Point Quiz Corrections
Drill & Practice Programs
Vincent Aleven & Kirsten Butcher
Simulated Student: Building Cognitive Model by Demonstration
Julie Booth, Robert Siegler, Ken Koedinger & Bethany Rittle-Johnson
Equations …. are mathematical sentences stating that two expressions are equivalent.
Presentation transcript:

SimStudent: A Computational Model of Learning as a Research Toolbox for the Sciences of Learning Noboru Matsuda Human-Computer Interaction Institute Carnegie Mellon University

Research Questions Building a cognitive model is hard. Can machine-learning techniques help non-expert authors build a Cognitive Tutor? Would like to simulate students’ learning. Can machine-learning techniques help us build a computational model of learning with a cognitive fidelity? I heard that students learn by teaching others. Can we use the computational model of learning to study the theory of learning by teaching?

Solution: SimStudent Machine learning agent Fundamental technology Learns procedural skills, by Observing model solutions & solving problems Fundamental technology Programming by Demonstration Inductive Logic Programming Knowledge representation Production rules (Jess) Lau & Weld (1998). Blessing (1997).

SimStudent Projects Intelligent Authoring Building a Cognitive Tutor as a CTAT Plug-in Student Modeling and Simulation Controlled educational studies Error formation study Prerequisite conceptual knowledge study Teachable Peer Learner Learning by teaching

Authoring a Cognitive Tutor Example-Tracing Tutor Little programming A cognitive model specific to a particular problem Limited generalization by editing a behavior graph Model-Tracing Tutor Powerful student model Cognitive task analysis is hard Writing production rules is even more challenging Performing a task is relatively easy…

Next Generation Authoring Build a tutor GUI Teaching a solution SimSt. learning Production Rules Rule simplify-LHS: IF is-equation( Eq ), is-lhs( Eq, Lhs ), polynomial( Lhs ), all-var-terms( Lhs ) Then simplify( Lhs, S-lhs ), enter( S-lhs ) Rule simplify-LHS: IF is-equation( Eq ), is-lhs( Eq, Lhs ), polynomial( Lhs ), all-var-terms( Lhs ) Then simplify( Lhs, S-lhs ), enter( S-lhs ) Rule simplify-LHS: IF is-equation( Eq ), is-lhs( Eq, Lhs ), polynomial( Lhs ), all-var-terms( Lhs ) Then simplify( Lhs, S-lhs ), enter( S-lhs )

Demo Authoring by Tutoring

Example: Learning to subtract a constant term Learning to subtract a constant number First example Subtract the difference between 4 and 3… subtract 1 Subtract the coefficient of X… I see 3x, 1, x, and 4 in the equation. I wonder where the ‘1’ came from… Subtract the last term on the left-hand side… PSLC Summer School 2012 :: SimStudent :: Noboru Matsuda (CMU)

Example: Learning to subtract a constant term Learning to subtract a constant number First example Subtract the difference between 4 and 3… subtract 1 Subtract the coefficient of X… I see 3x, 1, x, and 4 in the equation. I wonder where the ‘1’ came from… Subtract the last term on the left-hand side… PSLC Summer School 2012 :: SimStudent :: Noboru Matsuda (CMU)

PSLC Summer School 2012 :: SimStudent :: Noboru Matsuda (CMU) Prior Knowledge Feature predicates 18 predicates isFractionTerm(X), isConstant(X), isPolynomial(X),… Operators 42 operators add(X,Y), coefficient(X), getFrstNumber(X), … PSLC Summer School 2012 :: SimStudent :: Noboru Matsuda (CMU)

Example: Stoichiometry Tutor

SimStudent Projects Intelligent Authoring Building a Cognitive Tutor as a CTAT Plug-in Student Modeling and Simulation Controlled educational studies Error formation study Prerequisite conceptual knowledge study Teachable Peer Learner Learning by teaching

Model of Incorrect Learning Identify errors students commonly make Weaken SimStudent’s background knowledge Let SimStudent make an induction error

Weak Prior Knowledge Hypothesis Multiple ways to make sense of examples Get a coefficient and divide Get a denominator and multiply “multiply by x” 3x=5 “divide by 3” 4/x=5 “divide by 4” Get a number and divide

Results: Learning Rate Step Score Estimate of correct (probability of each step right) Precision -> step score # training problems Steps Score = 0 (if there is no rule applicable) # correct rule applications / # all rule applications

Student Model A set of knowledge components (KCs) (Li et al. EDM2011) A Machine Learning Approach for Automatic Student Model Discovery A set of knowledge components (KCs) Encoded in intelligent tutors to model how students solve problems E.g. How to proceed given problems of the form Nv=N One of the key factors that affects automated tutoring systems in making instructional decisions Previous Approach: Require expert input Highly subjective Proposed Approach: Use a machine-learning agent, SimStudent, to acquire knowledge 1 Skill  1 KC Skill application for that step  KC for each step

Human-generated vs SimStudent KCs Human-generated Model SimStudent Comment Total # of KCs 12 21 # of Basic Arithmetic Operation KCs 4 13 Split into finer grain sizes based on different problem forms # of Typein KCs Approximately the same # of Other Transformation Operation KCs (e.g. combine like terms) Arithmetic op are like “add to both sides”, subtract from both sides, etc. 4x = 20 vs. –x = 5

Results Significance Test Human-generated Model SimStudent AIC 6529 6448 3-Fold Cross Validation RMSE 0.4034 0.3997 Significance Test SimStudent outperforms the human- generated model in 4260 out of 6494 steps p < 0.001 SimStudent outperforms the human-generated model across 20 runs of cross validation AFM Performance proportion = 0.655990145 and odds 1.906893465 Can you compute separately this ratio for the steps where the Q matrix is different from

SimStudent Projects Intelligent Authoring Building a Cognitive Tutor as a CTAT Plug-in Student Modeling and Simulation Controlled educational studies Error formation study Prerequisite conceptual knowledge study Teachable Peer Learner Learning by teaching

Learning by Teaching SimStudent

Learn more about SimStudents Project Web www.SimStudent.org Contact us Noboru Matsuda mazda@cs.cmu.edu