+ All Categories
Home > Documents > LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

Date post: 05-Jan-2016
Category:
Upload: kellie-merritt
View: 214 times
Download: 1 times
Share this document with a friend
12
LInfoVis Winter 2011 Chris Culy Evaluation of visualizations
Transcript
Page 1: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Evaluation of visualizations

Page 2: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Levels: perception and interpetation

Visual/perceptual how well can people perceive the distinctions the

visualization intends?

Interpretation how well can people interpret the visualization?

Page 3: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Levels: Use

Use in isolation how accurately can people use the visualization for

a particular task in isolation? how quickly can people use the visualization for a

particular task in isolation? Use as part of a broader goal

how accurately can people use the visualization for a particular task as part of a broader goal?

how quickly can people use the visualization for a particular task as part of a broader goal?

Page 4: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Levels: Satisfaction

how satisfied are people with the visualization? e.g. easy/hard, "cool", etc. ≠ how well they can use it

how useful is the visualization for what they want to do? how well do they "get it"? people may use/prefer a more difficult tool

if it's "cool" if they already know it – there's a learning curve for a new

tool if it's cheaper

Page 5: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Goals

Formative evaluation Goal is to get specific information to improve the software Done during the development cycle Often informal

Summative evaluation Goal is to get general information about how the software

performs Done at the end of (a) development cycle Often (more) formal Can also be used to improve the next iteration

Page 6: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Some principles for experiments

There should be a specific purpose for the experiment What are you evaluating?

Experiments should be task based The user should be introduced to the software,

especially something new The user should not be “interfered with” during

the experiment

Page 7: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Some techniques for experiments

"instrumenting" the program

tracking clicks, etc, and then trying to analyse the patterns

also for tracking speed, accuracy

Page 8: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Some techniques for experiments

"instrumenting" the user various means for perception eye tracking

shows where people are paying attention

Page 9: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Some techniques for experiments

Observation with note taking, timing of the user's actions

User feedback pre/post session questionnaire "think aloud protocol": user explains what they're

doing and why as they're doing it explicit questions during demo (informal) post session interview (“debriefing”)

Page 10: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Formal vs. informal

Formal Several subjects (6+) Experimental design (comparison) Often statistical analysis of results Experiment usually recorded

Informal, "guerilla" testing Few subjects (3-4) Quick, informal test to get user feedback on s.t.

Page 11: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Techniques for deployed software

"Bug" reports When is a bug not a bug?

Feature requests

Interviews (or narratives) with "customers" Case studies (positive = “success stories”)

Observation of "customers" using the software

Page 12: LInfoVis Winter 2011 Chris Culy Evaluation of visualizations.

LInfoVis Winter 2011 Chris Culy

Interpreting users' reactions

Users can only react in terms of what they know/assume.

They don't analyse what they do, so their suggestions tend to be too concrete.

They often make specific suggestions that aren't necessarily the best solutions, since they don't know what the possibilities are

They are honestly trying to be helpful


Recommended