Help!

1
Help!

2
How do you get help when youre not using a
computer?
3
How do you get help when youre not using a
computer?

• From manuals and books
• By asking an expert
• By watching other people
• (librarian, receptionist, guard, etc.)
• Other people may notice and intervene
• (parents, tutors, coaches)

4
Getting help is quite a skill!

• You need to recognize youre having a problem.
• You need to know what you dont know.
• Forming the right question is hard.
• Knowing what to do with the answer is also hard.
• Getting help successfully requires quite a lot of
  metacognitive knowledge.

5
Forming the right question

• Question about how to ask a question
• How do I search for a line in a file?
• The right question
• man grep
• The answer
• grep lttargetgt ltfilenamegt
• (ex. from UNIX)

6
Some ways to classify help

• Is it personalized or context-sensitive or
  generic?
• Is it provided for vague problems? (huh?) Or
  must a request be specific?
• Initiative is help offered, or must it
  requested?
• Source human, document, computer?

7

• A good on-line help system is much more than just
  the manual, on-line.
• A good help system might take into account
• a users characteristics
• what shes trying to do
• the context shes trying to do it in

8
Computer help systems

• On-line or off-line
• Context sensitive or not
• Browsable
• Searchable
• Generic

9
Examples of on-line help

• Back-Seat Driver
• Hypertext (invented by Ted Nelson)
• Error messages
• Undo
• Did you know..?
• User modeling
• Grundy (stereotyping users)
• Eager (observing users)

10
Approaches to user modeling

• Stereotype the user
• Grundy, MIT movie manuals
• Observe the users behavior
• Eager - automatic macros (Allen Cypher)
• Amazon and other programs that make
  recommendations
• Issue if the system notices something, how much
  initiative should it take?

11
Study of on-line tutorials

• Subjects 10th and 11th grade boys
• Task Create a database of favorite songs
• 4 different help conditions
• General-to-specific
• Explanation-to-specific
• Specific-to-specific
• Explicit instructions
• Measures learning test time, errors

12
Results

• Learning time test time errors
• General 2059 640 7
• Explanation 2315 744 7.7
• Specific 2635 725 10
• Explicit 1648 1009 9.9

13
Moores study of explanations

• Analyzed what experts do
• Respond to huh
• Produce clarifications in response to indications
  of misunderstanding
• Explanation is not a one-shot deal

14
Moores conclusions

• Experts should know a lot about users
• Focus on feedback, not on stereotypes
• Keep track of context
• Be able to illustrate the same point in many ways
• A reactive model of explanation is best

15
Conclusions problems with help

• When the user has initiative
• Questions can be difficult to frame.
• Mapping problems onto the systems options may be
  difficult.
• When the system has the initiative
• Error messages usually ignore context.
• It may offer help the user doesnt want.