ACT-PRO Action Protocol Tracer A Tool for Analyzing Simple, Rule- based Tasks Wai-Tat Fu & Wayne D. Gray ARCH Lab George Mason University.

Slides:



Advertisements
Similar presentations
Chapter 14: Usability testing and field studies. 2 FJK User-Centered Design and Development Instructor: Franz J. Kurfess Computer Science Dept.
Advertisements

©N. Hari Narayanan Computer Science & Software Engineering Auburn University 1 COMP 7620 Evaluation Chapter 9.
ECE Synthesis & Verification1 ECE 667 Spring 2011 Synthesis and Verification of Digital Systems Verification Introduction.
DDMLab – September 27, ACT-R models of training Cleotilde Gonzalez and Brad Best Dynamic Decision Making Laboratory (DDMLab)
Structural Bioinformatics Dr. Avraham Samson Course no.: Credit points: 1.5 Final grade is based on 10 assignments Course homepage:
GOMS and keystroke predictive methods Judy Kay CHAI: Computer human adapted interaction research group School of Information Technologies.
Predictive Evaluation Predicting performance. Predictive Models Translate empirical evidence into theories and models that can influence design. Performance.
Winter Retreat Connecting the Dots: Using Runtime Paths for Macro Analysis Mike Chen, Emre Kıcıman, Anthony Accardi, Armando Fox, Eric Brewer
Bottom-Up Integration Testing After unit testing of individual components the components are combined together into a system. Bottom-Up Integration: each.
Statistical Natural Language Processing. What is NLP?  Natural Language Processing (NLP), or Computational Linguistics, is concerned with theoretical.
User Interface Design Chapter 11. Objectives  Understand several fundamental user interface (UI) design principles.  Understand the process of UI design.
Adaptive Signal Processing Class Project Adaptive Interacting Multiple Model Technique for Tracking Maneuvering Targets Viji Paul, Sahay Shishir Brijendra,
Machine Learning in Simulation-Based Analysis 1 Li-C. Wang, Malgorzata Marek-Sadowska University of California, Santa Barbara.
Methods for checking simulation correctness How do you know if your testcase passed or failed?
From Devices to People: Attribution of Search Activity in Multi-User Settings Ryen White, Ahmed Hassan, Adish Singla, Eric Horvitz Microsoft Research,
Chapter 5 Models and theories 1. Cognitive modeling If we can build a model of how a user works, then we can predict how s/he will interact with the interface.
Assignments  1. Grade graphs and conclusions.  2. Introduction to Reaction Time.  3. Begin Pre-Lab of Reaction Time.
Contextual Inquiry A fact-based approach to understanding the reality of users’ goals, processes, and tasks. Puts designers, engineers, researchers, etc.
Overview of the rest of the semester Building on Assignment 1 Using iterative prototyping.
Online Marketing Messages Description / guideline Ronni Hartvig November 2010.
COMP5047 Pervasive Computing: 2012 GOMS and keystroke predictive methods Judy Kay CHAI: Computer human adapted interaction research group School of Information.
(Business) Process Centric Exchanges
Testing & modeling users. The aims Describe how to do user testing. Discuss the differences between user testing, usability testing and research experiments.
Identifying needs and establishing requirements
GOMs and Action Analysis and more. 1.GOMS 2.Action Analysis.
SUPERCOMPUTING CHALLENGE KICKOFF 2015 A Model for Computational Science Investigations Oct 2015 © challenge.org Supercomputing Around.
A Model for Computational Science Investigations Supercomputing Challenge
Creating Graphical User Interfaces (GUI’s) with MATLAB By Jeffrey A. Webb OSU Gateway Coalition Member.
FIX Eye FIX Eye Getting started: The guide EPAM Systems B2BITS.
Cognitive Modeling / University of Groningen / / Artificial Intelligence |RENSSELAER| Cognitive Science CogWorks Laboratories › Christian P. Janssen ›
Copyright 2006 John Wiley & Sons, Inc. Chapter 1 - Introduction HCI: Designing Effective Organizational Systems Dov Te’eni Jane Carey Ping Zhang.
Chapter 10 Verification and Validation of Simulation Models
ITM 734 Introduction to Human Factors in Information Systems
The word science comes from the Latin "scientia," meaning knowledge. Scientific Theories are not "tentative ideas" or "hunches". The word "theory" is often.
Computer Science 1 Test Selection and Augmentation of Regression System Tests for Security Policy Evolution JeeHyun Hwang, Tao Xie, and collaborators at.
OPERATING SYSTEMS CS 3530 Summer 2014 Systems and Models Chapter 03.
U i Modleing SGOMS in ACT-R: Linking Macro- and Microcognition Lee Yong Ho.
Dynamic Decision Making Laboratory Carnegie Mellon University 1 Social and Decision Sciences Department ACT-R models of training Cleotilde Gonzalez and.
Computer Simulation of Networks ECE/CSC 777: Telecommunications Network Design Fall, 2013, Rudra Dutta.
Copyright 2006 John Wiley & Sons, Inc. Chapter 1 - Introduction.
Chapter 1 - Introduction
1CS 338: Graphical User Interfaces. Dario Salvucci, Drexel University. Lecture 15: User Modeling.
Software Quality Assurance and Testing Fazal Rehman Shamil.
TRUSTED FLOW: Why, How and Where??? Moti Yung Columbia University.
Understanding Naturally Conveyed Explanations of Device Behavior Michael Oltmans and Randall Davis MIT Artificial Intelligence Lab.
GOMS as a Simulation of Cognition Frank Ritter, Olivier Georgeon 28 oct 2014.
Chapter 11 user support. Overview Users require different types of support at different times. There are four main types of assistance that users require:
Method and Application of Protocol Analysis Anguo Dong Computer Science Department, University of Calgary.
Jeremy Nimmer, page 1 Automatic Generation of Program Specifications Jeremy Nimmer MIT Lab for Computer Science Joint work with.
1 Decision Making ADMI 6510 Simulation Key Sources: Data Analysis and Decision Making (Albrigth, Winston and Zappe) An Introduction to Management Science:
1 Testing Implementations of Access Control and Authentication Graduate Students: Ammar Masood K. Jayaram School of Electrical and Computer Engineering.
Cognitive Informatics for Biomedicine – Chapter 5
FOP: Multi-Screen Apps
OPERATING SYSTEMS CS 3502 Fall 2017
Human Computer Interaction
Security Testing Methods
Chapter 1 - Introduction
Introduction to Event-Driven Programming
E-learning. DongFeng Liu, Priscila Valdiviezo, Guido Riofrío, Rodrigo Barba November 19, 2015 VARE dfgsdgfsgdsgsdgaga.
Chapter 10 Verification and Validation of Simulation Models
Computer Simulation of Networks
Test Case Purification for Improving Fault Localization
“In the midst of chaos, there is also opportunity” - Sun Tzu
MyFloridaMarketPlace
GOMS as a Simulation of Cognition
GOMS as a Simulation of Cognition
Cognitive models linguistic physical and device architectural
Testing & modeling users
Chapter 1: Creating a Program.
Presentation transcript:

ACT-PRO Action Protocol Tracer A Tool for Analyzing Simple, Rule- based Tasks Wai-Tat Fu & Wayne D. Gray ARCH Lab George Mason University

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks2 Overview Problem What is ACT-PRO? Utility of ACT-PRO How it works

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks3 *where correct = behavior that the model knows about Problem Model tracing correct* behavior –Successfully developed and used to shadow students as they perform a task –Usually combined with tutors that keep students from straying off the correct path Can model tracing techniques be applied after-the-fact to large quantities of action protocol data? Model tracing errors, error detection, and error correction –Usually done “by hand” or not at all Can the tracing of errors, error detection, and error correction be handled as routinely as the tracing of correct behavior is handled by existing tutoring systems?

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks4 ACT-PRO Developed to handle a class of interactive tasks –Programming VCR, flight computer, ATM, etc Requires task analysis of interactive behavior –Format of task analysis is irrelevant; e.g., ACT-R, GOMS, Soar, or Hierarchical Task Analysis Currently implemented in visual basic for applications

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks5 The Task -- Overview ACT-PRO has been applied to action protocols collected from Ss programming a simulated VCR The VCR is representative of a large class of human- computer interaction tasks (e.g., ATMs, flight computers, and so on) The ACT-PRO approach should work with this class of interactive tasks

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks6 Level of Effort How does ACT-PRO compare to alternative methods of classifying action protocols? ≈ 12 days to program ≈ 12 min to setup and run ≈ 12 hr to verify results

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks7 Overview of Rest of Talk The task in more detail ACT-PRO -- how it works

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks8 The Task A simple rule-based task - programming a simulated commercial VCR

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks9 Device-Task Rule-Hierarchy Example: Start Time: 7:00; End Time: 8:00; Dow: Sat Channel: 11

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks10 Focus of the Study Focus of the study is on the explanation of human errors in terms of the model’s control structure By comparing the human data with the model’s prediction, we can characterize human behavior as matching or violating the model’s behavior ACT-PRO is needed to correctly categorize the correct behavior and to classify errors, error detection, and error correction

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks11 ACT-PRO

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks12 Matching Protocols to Model HOUR 10MIN MIN Push-goal-SET-StartHour Click On HOUR Pop-goal-SET-StartHour Push-goal-SET-Start10Min Click On 10MIN Pop-goal-SET-Start10Min Push-goal-SET-StartMin Click On MIN Pop-goal-SET-StartMin Button Clicks from Ss Model

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks13 ACT-PRO Two levels: –Keypress level Classify each button click as matching or violating the model –Goal level Parse button clicks into goal pushes and pops Classify each goal push and pop as matching or violating the model

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks14 Keypress Level Button clicks are classified as a type of hit or error* A transition matrix is used to parse the button clicks Within context defined by prior button clicks, clicks are unique to a certain goal * see Gray, W. D. (in press). The nature and processing of errors in interactive behavior. Cognitive Science... Btn 2 Btn 3 - hit Btn 1 - error..

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks15 Goal Level Parse button clicks into goal pushes and pops Classified each push and pop as a hit or error –Hits are labeled according to the ideal model (i.e., ACT-R productions or other task analysis) –Errors are further classified into: Push-error: Ss attempted to do something that the model would not do. Pop-error: Ss stopped working on a goal either before or after the model would stop.

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks16 Example 1 - Push-error Show Time End Time: 20:00 ButtonDisplay END-MODEE:EE HOUR19:00 HOUR20:00 10MIN20:10 10MIN20:20 10MIN20:30 10MIN20:40 10MIN20:50 10MIN20:00 DOW20:00 Error Hit Error Push-error Error detection and correction

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks17 Example 2 - Pop-error Show Time Start Time 21:15 ButtonDisplay START-MODEE:EE HOUR21:34 10MIN21:44 MIN21:45 10MIN21:55 10MIN21:05 10MIN21:15 Pop-error Goal not accomplished yet. Error detection and correction

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks18 Results KeypressGoal push/pop Total Correct Behavior80.1%94.81% Errors, Error Detection, or Error Correction 19.9%5.19% Correctly Classified by ACT-PRO 98.95%99.30% VCR Condition A: Ubiquitous Information

Fu & Gray ACT-PRO -- A tool for simple, rule-based tasks19 Conclusions ACT-PRO works very well Involves several orders of magnitude less effort than hand encoding Well suited to modeling behavior from a large class of interactive devices Next challenge: –Condition B - where target settings are covered by gray boxes and Ss need to click on the gray box to access the information –Condition C - Same as Cond B, but Ss are forced to memorize the target settings before the experiment