1. Adaptive
Interfaces
Jeffrey Heer · 28 May 2009

2. Final Project Presentations
Tuesday June 9, 3:30-6:30pm, 104 Gates
8 minute presentations
6 min for research, 2 min for questions
Start with an overview:
1 sentence statement of your research result
1 slide / 4 sentences of what you did and why
Rest of time on details. Assume audience is familiar with HCI issues: focus on your work

3. Final Project Participation
Be sure to fill out the survey accompanying Sharon’s paper reader! Do others want to ask the class to participate in their research?

4. Direct Manipulation vs. Agents!
Public Debates at CHI ‘97, IUI ‘97 Ben Shneiderman for Direct Manipulation Pattie Maes for Interface Agents

5. Ben Shneiderman
“I think we would do best to focus on the remarkable human capabilities in the visual domain, which I think are largely under-utilized by the current designs with 40 icons in 2-3 windows. I think we should have two or three orders of magnitude more: 4,000 or more items on the screen in an orderly way that enables people to see all of the possibilities and navigate among them.”

6. The Direct Manipulation Ideology
Display as much information as possible
Predictable
Rapid, reversable interactions
User initiates all actions

7. The goal: high information density

8. Command Line: Low density and indirect manipulation

9. guis can provide improved density and more direct manipulation…

10. …but still have a ways to go

11. Pattie Maes
“Why do we need software agents? … Take a look at the World Wide Web, for example. You couldn’t possibly try to visualize the World Wide Web in any way because it is completely unstructured and because it has been built by so many different people and is continuously changing. I believe that the dominant metaphor that we have today is a mismatch for the computer environment we are dealing with tomorrow.”

12. The Intelligent Interfaces Ideology
Agents know habits, preferences, interests
Tasks can be delegated to software agents
Mixed-initiative: computer can be proactive
prompt-based telephone interfaces are an example of complete computer initiative

13. Some Successes Spam Filtering Collaborative Filtering
Toyota Prius power train and braking

14. Failures?
From

15. DM v. Collaborative Filtering

16. Pros and Cons?
Predictability and Intelligibility Personalization and Adaptation Reactive vs. Proactive Scalability, Accuracy, Anthropomorphism

17. Adaptive Interfaces Build model of user and/or context (device)
Preferences
Abilities
Build optimization / selection criteria
Adapt interface design or take action
Perform optimization/search with model criteria
Change layout, highlighting, exposed features
Recommend selected options / items
(Sometimes) make decisions on when to act

18. Menus – Does Adaptation Help?
Based on usage, try accelerate access
Options
Cull unlikely menus
Split the menu into predicted and normal
Ephemeral adaptation
Why might this fail?

19. Collaborative Filtering
Recommend based on similarity to others
Herlocker, J. L., Konstan, J. A., Terveen, L. G., and Riedl, J. T Evaluating collaborative filtering recommender systems. ACM Trans. Inf. Syst. 22, 1 (Jan. 2004), DOI=

20. Collaborative Filtering
Recommend based on similarity to others
How to evaluate?
Accuracy measures (mismatch with user ratings)
Novelty, “non-obviousness”
Coverage – how much should be recommended?
Algorithms vary based on data (# users vs. # items)
Inherent variability: people’s ratings may not be consistent over multiple samples
Herlocker, J. L., Konstan, J. A., Terveen, L. G., and Riedl, J. T Evaluating collaborative filtering recommender systems. ACM Trans. Inf. Syst. 22, 1 (Jan. 2004), DOI=

21. Supple [Gajos et al]

22. User Preference Elicitation

23. Model Motor Abilities

25. Evaluation

26. Personal Universal Controller [Nichols et al]

27. Personal Universal Controller
How to deal with multiplicity of devices?
How to control ubicomp environments?
Specification language
Specify complete functionality
of appliances
Use specifications to generate
interfaces for mobile devices

28. Ensuring Interface Consistency

29. Trade-offs for Generated UIs
Personalized to preferences / abilities
Cross-device functionality
Poor aesthetics
Only valuable when human designer absent?

30. CS547 Tomorrow
Designing Online Communities from Theory Robert Kraut, CMU HCI Institute 12:30-2pm, Gates B1