comp5047 pervasive computing: 2012 think-aloud usability experiments or concurrent verbal accounts...
TRANSCRIPT
COMP5047 Pervasive Computing: 2012COMP5047 Pervasive Computing: 2012
Think-aloud usability experiments or concurrent verbal accounts
Judy Kay
CHAI: Computer human adapted interaction research group
School of Information Technologies
Postconditions for this week
• Describe the uses of Think Alouds• Describe the processes for conducting one• Describe advantages and limitations
• Ability to use Think-aloud as relevant and for your project
• Justify the use of Think Aloud in the overall testing of a pervasive computing application
• Be ready to do Think Aloud evaluations for your part of your project
Think aloud protocols
• Ask user to “think aloud” as they use the interface
• Often used with video, audio taping• Otherwise MUST make notes
• Helps observer interpret what is going on• Gives qualitative data mainly• 3-5 users may be enough (Nielsen) for
each stage of refining the prototype
Case study
• Test usability of teacher’s data projection facilities in this classroom
• What is intended functionality?• Formulate some concrete tasks – write
these as instructions– eg Suppose you currently have the projector
screen down but you want to use the whiteboard. So you want make the screen go up.
Case study
• Work in your groups to identify 3 tasks relevant to testing the teacher’s use of projection facilities in this room
Case study
• Call for volunteers to be users for this trial• Have not used these facilities before
• Have used them in other places in the uni
• Call for volunteers to conduct the trial• All will make helpful notes
Classification of think-aloud
• formative versus summative• predictive v empirical• laboratory v naturalistic• qualitative v quantitative
• For the project, you goal will be?
Design cycle
• User Centred Design• Define criteria for success• Define concrete tasks users should be able
to do - use these in evaluation• Prototype construction• Usability study
– Decide just what data to collect– Test design of experiment for timing (trial it)– Recruit users– Run study
• Goto top
Recruiting users
• How representative are they?– similarity to intended user population
– Age
– Gender
– experience in area
– interest/motivation
– computer literacy
• What effect does user population have for conclusions?
Stages of running an evaluation
1. Preparation
2. Introduction
3. The test
4. Questionnaire/interview
5. Debriefing
6. Analysis, reflection, summarising, reporting, conclusions for action
Steps 1- 5 done for each user test, as run
Step 6 is applied mainly after several users
Preparation
• Set up machine, room, environment• Check all of them• Check user instructions• Do a mental run-through
Be sure not to waste user's time because of your lack of preparation!
Introduction
• Welcome user, explain purpose of test– make clear system tested not user
– confidentiality
– anonymity of reporting
– opt out at any time
– what is recorded
• Invite any other questions to here– explain procedure
– if appropriate, do demo
– invite questions
The test
• User works through experiment....– recording
– ensure user feels supported
– show pleasure at problems identified
– critical to help user if stuck
• Questionnaire/interview– open and closed
What data should you collect?
• Observe – direct/indirect
– take notes
– video/audio/software monitor
– software logs for timing
• Questionnaire: – open
– closed
Debriefing
• Thank user• Remind them of usefulness of results• Pause to make sure all data collected• All notes written• May ask user to confirm details collected
Pitfalls
• Defining the right concrete tasks– Test all key aspects
– Multiple tasks for same aspects
• Instructions to the users– Do NOT lead the user
– Take particular care not to use words that are identical to terms on the interface
Benefits of think aloud
• “show what users are doing and why they are doing it while they are doing it in order to avoid later rationalisations”
(Nielsen, Usability Engineering, Academic press 1993, p195)
• Cheap• Slows users down
– studies show users may work faster with fewer errors due to care on critical elements
Problems of think aloud
• Not directly quantitative
• Add cognitive load to users
• User's “theories” must be interpreted with care
• Slows users down
• Users are aware they are being observed so behave accordingly
Facilitating think aloud
• What are you thinking now?• What do you think that message means?
(only after the user has noticed the message and is clearly spending time on it)
• don't help user except withHow do you think you can do it?• if user appears surprised, Is that what you expected to happen?
Naturalised think-aloud
• Multi-user interaction– Two (or more) users work on task
– Conversation is natural
– Observer collects dialogue
Problems with observing users
• Hawthorn effect
• People rationalise (Telling more than we can know)
• Qualitative data
Class activity - foreshadowed
• Formulate three concrete tasks for testing the console of the lecture theatre
• Write these out ready for use in an experiment
• After next part of the seminar: each group will conduct one practice Think-Aloud with the rest of the class writing feedback on how this was done.