Presentation is loading. Please wait.

Presentation is loading. Please wait.

Planning who, what, when, and where

Similar presentations


Presentation on theme: "Planning who, what, when, and where"— Presentation transcript:

1 Planning who, what, when, and where
Chapter 22 Planning who, what, when, and where

2 Intro We have da strategy: Now: purpose type of data to collect system
constraints Now: choose users (population sample) create timetable (and stick to it) prepare task descriptions (script it!) decide where to evaluate (field or lab)

3 Choosing participants
Each participant should be: a real (actual) user, or representative user (from requirements), or usability or domain expert Participants should not be: chosen at random (hmm, this is contrary to “traditional” experimental criteria…why?)

4 Screening or pretesting
To gage whether user fits desired subject profile, may need to screen users For example, if testing a Spanish language training program, don’t use fluent Spanish language speakers Questionnaires may be used to record users’ experience levels (can be useful in analysis later, e.g., discard experts’ data)

5 Examples System Actual users Participants (early prototype)
Public info tourist kiosk The public: possibly foreign (lacking language skills) Actual users, but English ok; fix early errors first Safety critical system for monitoring car assembly line Plant supervisors with job experience Actual users due to importance of domain knowledge Web site offering tax filing services (e.g., TurboTax) People who want to file their taxes Actual users of desktop systems

6 Working alone or in Pairs?
Usually users are tested alone When may pairs be a good idea? working cooperatively, sharing a computer different culture (e.g., Japanese) they prefer it (e.g., husband/wife team) Hire a facilitator / caretaker / custodian? when working with children, disabled, etc. when interpreter is needed

7 How many participants? Depends on problem and stage of testing
“trivial” or “easy” troublespots will be identified quickly (and often if many users); so need only a few during early stages OTOH, if a couple of users finds no problems, does that mean UI is acceptable in general? Key issue: generalizability Ideally, you’d want to conduct a power analysis of the experiment But one typically goes with “rule of thumb”: start with 5, go to 10, etc………………..

8 University participants
The book’s discussion seems to be aimed at practitioners What about at the Uni? use the Psych pool, other students, etc. problems: restricted age group users may not have required expertise (e.g., evaluating a Fortran debugger) motivation may be wanting (e.g., doing it for credit, not so much for science or “good of humanity” :)

9 Incentive If possible, compensate users: extra credit
real credit (e.g., on an e-commerce web site) food nick-nacks (mugs, pens, bla bla) soap (true story :) money is always good, if you have enough to spare

10 Global Warming App Users
How they picked users for the Global Warming App study: solicitation experienced users (not novices) various disciplines (e.g., not CS necessarily) 10 users no incentives

11 Create a timetable Timetable:
How long do you need per evaluation session? may need to run quick pilot study to determine How much time will the whole process take? quick “back of the envelope” calculation: 100 subjects, 10 minutes each = 16 hours (not counting introductions, filling out questionnaires, lunch, dinner, classes, interruptions, Survivor episodes, etc.) how many sessions can you run per day? Maybe 4-5 hours’ worth? so what’s a realistic estimate for 100 subjects? week? two weeks?

12 Timetable (cont.) Keep evaluation session to a minimum, try not to exceed 1 hour (subjects will get bored, tired) Create a timetable “sign up sheet” very useful for signing up subjects and reserving lab space (e.g., eye trackers) Allocate time for analysis 80% of time spent in analysis (true? I dunno, just guessing) just like debugging code?

13 Task Descriptions Create task descriptions
similar to idea of scripts for evaluators (so they know what to say and say the same thing to each participant, thereby reducing bias) these are scripts for users I think they’re a good idea (task cards), but so long as they’re not too detailed case study: my Navy usability study: participants read directions from a script. This was too easy; everyone performed similarly; no clear performance problems were identified

14 Where to do evaluation? Field studies Controlled studies
observations in the field: most realistic environment, obviously lacks control Controlled studies in a lab, usually or some kind of mock scenario (e.g., “shoot house” for cops, SWAT personnel)

15 Usability Lab How to build a good usability lab:
often a separate room is used for participants (one-way mirror) various logging devices, e.g., cameras, keystroke logging software, etc. do we have one at Clemson? We should…


Download ppt "Planning who, what, when, and where"

Similar presentations


Ads by Google