1 / 29

Overview

Stereotypes, student models and Scrutability Judy Kay Basser Department of Computer Science Madsen Building, F09 University of Sydney NSW 2006 AUSTRALIA email judy@cs.usyd.edu.au. Overview. Stereotypes Pervasive Natural um community but not ITS/AIED Scrutability and student models

henrygordon
Télécharger la présentation

Overview

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Stereotypes, student models and ScrutabilityJudy KayBasser Department of Computer ScienceMadsen Building, F09University of Sydney NSW 2006AUSTRALIAemail judy@cs.usyd.edu.au

  2. Overview • Stereotypes • Pervasive • Natural • um community but not ITS/AIED • Scrutability and student models • Stereotypes and scrutability

  3. Example: Stereotype for ITS’2000 attendee Friday 9am • Expert in education • Expert in AI • Stayed up too late last night • Saturated with great new insights in ITS • Happy with `intelligent’ in ITS • Unaware of Fiji coup

  4. More…….. • Have not read my paper in the proceedings • Very interested in student modelling • Had not heard term scrutability before this week • Unconvinced of merit of scrutability of student/learner models • Unconvinced of merit of stereotypes

  5. Classic stereotypes • Beginner vs advanced • Default is beginner • Stereotypes about beginners • A little initial information about the student allows inference of a rich initial default student model • May well be very rough (cf nothing) • ?

  6. Rich: A stereotype represents a collection of attributes that often co-occur in people. ...they enable the system to make a large number of plausible inferences on the basis of a substantially small number of observations. These inferences must however, be treated as defaults which can be overridden by specific observations. Example: athletic person motivated by excitement (0.7) strength and perseverance (0.8) interested in sports (0.63) Elements: • Initial interaction • Activate stereotypes • DAG of stereotypes • On-going refinement

  7. Double stereotypes(Chin, 1989) • Reason from user actions • Observed actions • Assume actions known • Infer expertise classification • Reason from expertise classification • Infer huge number of actions known

  8. Tightening the Stereotype? {trigger function activate stereotype} {retraction function deactivate stereotype} {failed essential trigger deactivate stereotype} active stereotype Many inferences • Triggers • Active stereotype v inactive stereotype • Retraction condition • Essential triggers • High fanout of inferences

  9. Statistical Character • Validity is statistical (ie not individual) • Threshold probability for each inference Can be based upon empirical evidence: Active (expert) Knows (reg exp) 0.87

  10. Knows (A) Knows (loops) Knows (variables) (0.99) What is not a Stereotype? B prerequisite of A Eg. Know loops (0.6)

  11. Stereotype and classic student modelling • Overlays • ‘expert’ • Differential models • ‘plausibly ideal student’ • Buggy models • ‘classic errors’

  12. Stereotype for ITS’2000 attendee Friday 9.22am • Expert in education • Expert in AI • Stayed up too late last night • Saturated with great new insights in ITS • Happy with `intelligent’ in title ITS

  13. More…….. • Very interested in student modelling • Had not heard term scrutability before this week • Unconvinced of merit of scrutability of student/learner models • Unconvinced of merit of stereotypes

  14. Building Stereotypes • Handcrafted stereotypes • Eg beginner, advanced • Local stereotypes • Empirically-based stereotypes • Machine learning • Statistical analysis • Cliques, communities

  15. Scrutability A Scenario Jonathan uses a text editor He does not know about its undo. A coach tells him about it, why it is useful and how to use it. Jonathan tries it. He likes it.

  16. What Jonathan might like to know? How did coach know that I didn’t know about undo? What else does coach think I know? Or don’t know? Why did coach tell me about undo? How can I tell coach what I want to know? Why did coach explain undo that way? Does coach explain it differently to other people? Is this the sort of inquisitive student you want? Should we want students to question, explore, …?

  17. sam and the Basser Study (Kay and Thomas, CACM 1995) • Monitor data as basis for building scrutable student models • Data on growth of expertise • 10 years, ~600 new users per year initially • Field trial of users in their 4th semester

  18. go_k scroll 2 minimal scroll_1_3 basics no_scroll_b

  19. go_k scroll 2 minimal scroll_1_3 basics no_scroll_b Go_k Show evidence Explain Set value to: True Set value to: False Set value to: Maybe

  20. Explanation of go_k Clicking anywhere in the current window moves the cursor To that place in that window. Start typing and characters Appear right of the cursor. Clicking on a window other the current window makes that new window (and that file/buffer) current.

  21. Explanation of go_k Once you have selected a window with one click on Button 1, the text you type goes where the cursor is. To move the cursor in the current window, click button 1 Where you want to type. For example, if you type `hollo world’ when you meant to type `hello world’, you click (with button 1) at the point between the `o’ and the `l’ in `hollo’, type the backspace key and an `e’, then click again after `there’ to continue typing where you left off.

  22. Scrutability • Ontological • Values of student model components • Bases • Reasoning mechanisms • Big student models

  23. Stereotypes and Scrutability • Am I a beginner? • What are the implications of being a beginner? • What would be different if I were an expert? • How can I let the system model me as a beginner, but have it recognise some of the more advanced things I know?

  24. Buggy Stereotype as Learning Objects Scenario One Beginner Python programmer tries to write a program. Student hits a problem. ITS diagnoses the difficulty. Scenario Two Beginner Python programmer tries to write a program. Student hits a problem. Student consults list of stereotypic errors. (For beginners in this task, at this stage).

  25. Stereotype Companions eg Hietala Niemirepo 1998 • Young children learning mathematics • Several artificial learning companions • Able (stereotype?) • Weak (stereotype?)

  26. Conclusions on scrutability of student models • Reflection • Learner control and responsibility • Metacognition • `Correctness’ of model • Humans may be inscrutable to machine but • Machines and people are different • Vive la difference!

  27. Conclusions on stereotypes and scrutability • Triggers (essential and other) • Large fanout of inferences • Learner control of probability allowed • Tuning selected inferences • Activation and retraction • User control

  28. Final conclusions • ITS’2000 attendees at 9.50 • Feel they knew scrutability all their lives • Believe scrutable student models may be ok • Feel familiar with the notion of stereotypes in student models • Are likely to think about scrutable stereotypes in their own systems • Are very anxious for morning tea

  29. Thank you!Over to youfor questions

More Related