is your user hunting or gathering insights? identifying insight drivers across domains
DESCRIPTION
Presenter: Michael Smuc, Eva Mayr, Hanna RiskuBELIV 2010 Workshophttp://www.beliv.org/beliv2010/TRANSCRIPT
Is Your User Hunting or Gathering Insights? Identifying Insight Drivers Across Domains
Michael Smuc, Eva Mayr, Hanna Risku
Motivation
Goal: Evaluate a visualization tool for temporal pattern analysis
Expert-User DilemmaReal domain experts are rare, hard to find (and to motivate)Sometimes they don‘t even exist
Across-domain developmentWhat if you wanted to develop a tool suitable for different domains?
Solutions
Use heuristics instead of insight analysisIn case you find suitable ones for your tool, you can‘t keep in touch with the users
Educate non-experts time consuming, „they will tell you what you taught them to tell” even more time-consuming for multiple domains
=> we would like to propose another solution
Our approach
► Across-domain testing: Experts have to solve tasks in data visualizations of their own domain and from other domains
► Additional aid story about the data
► Research questions How do experts differ in different domains? More insights?Results only useful when experts work in own domain?
… and what makes an domain expert?
Specific knowledge of domain experts
Common ground of domain experts in our case „temporal data explorers“
Insight Study
► 9 experts in temporal data analysis from 4 different domains
► think-aloud
► Insights: quantitative & qualitative analysis
Q: Effect of domain expertise?
05
101520
25303540
time on task(min)
sec/datainsight
number ofoverview
data insights
number ofdetail data
insights
domain expertise no domain expertise
t = -0.29, df = 24,
p > .05
=> Across domain testing works!
Effect of domain expertise on insights
Different kinds of insights in a catering business dataset
Type 1: "At noon there is a red belt." Type 2: "There is quite some breakfast business."
Type 1
Type 2
Typology
► Type 1: insight-gatherers: simple description
► Type 2: insight-hunters: active search for insights, driven by prior knowledge
“…. even the smallest domain information is used to create novel interpretations and make as much sense of the data as possible”
no sign. differences for the number of insights but significant differences for use of prior
knowledge & hypotheses
Is insight hunting driven by domain-expertise?
NO, 50% of the domain experts, but 25% of the non-domain experts showed this behavior (no sign., but small dataset)
only every second domain expert hunted for insights
Discussion
► Across-domain-testing works, but domain expertise needs a differentiated approach? (redefinition?)
► experts common ground & the story are insight-drivers
► Even shallow insights are useful, insights interrelate
► Typology iHunters | iGatherers
allows selective sampling
“hunters’ insights are the best argument to sell“
sampling is easier
“sometimes it is sufficient to gather the second choice experts”
Questions?
what is expertise?
across domain testing
shallow insightsRelational Insight Organizer RIO
what is a domain?insight hunting
insight gathering
experimental setting
story about the data
sampling
what makes an expert?prior knowledge
compensation by experts common ground
applicability
future research
generalizability
additional
0510152025
data insights
hypotheses
prior knowledge
data insights
hypotheses
prior knowledge
ed
uca
tio
nfi
na
nce
insight hunters insight gatherers
t = 0.16, df = 14, p > .05
t = -3.80, df = 7, p < .01
t = -4.87, df = 14, p < .001
Definition of insights
“… the understanding gained by an individual using a visualization tool (or parts thereof) for the purpose of data analysis, which is a gradual process towards discovering new knowledge“