Analysis and Validation - Lecture 4 - Information Visualisation (4019538FNR)

This lecture forms part of the course Information Visualisation given at the Vrije Universiteit Brussel.

Transcript

1. 2 December 2005 Information Visualisation Analysis and Validation Prof. Beat Signer Department of Computer Science Vrije Universiteit Brussel beatsigner.com
2. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 2 March 10, 2022 Four Nested Levels of Vis Design
3. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 3 March 10, 2022 Validation ▪ Huge vis design space and most designs are ineffective ▪ Validate choices right from the beginning of the design process ▪ top-down design (problem-driven work) - start at top situation domain level - major challenge at data/task abstraction; mainly use existing idioms ▪ bottom up design (technique-driven work) - invention of new idioms or algorithms ▪ Validate all four levels of the design ▪ domain validation ▪ abstraction validation (what and why) ▪ idiom validation (how) ▪ algorithm validation
4. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 4 March 10, 2022 Four Nested Levels of Vis Design ▪ Output from upstream level is input to downstream level ▪ errors at upstream levels propagate to downstream levels ▪ highly iterative design process
5. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 5 March 10, 2022 Domain Situation ▪ A domain situation is defined by ▪ target users ▪ domain of interest of target users - each domain might have its own vocabulary ▪ data of target users ▪ questions (tasks) of target users ▪ Outcome of design process ▪ understanding of user needs (user-centred design) - e.g. via observations or interviews ▪ Challenges and risks ▪ users can often not clearly specify their analysis needs ▪ designers make assumptions (rather than engaging with users)
6. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 6 March 10, 2022 Data & Task Abstraction ▪ Abstract from answers to domain-specific questions at upstream to a generic representation ▪ questions from different domain situations can map to same abstract vis tasks - e.g. browsing, comparing or summarising ▪ Design abstract data ▪ data from upstream is often transformed to something different ▪ determine which data type supports a visual representation that addresses a user's problem
7. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 7 March 10, 2022 Visual Encoding & Interaction Idiom ▪ Specific way (idiom) to create and manipulate the visual representation of abstract data ▪ visual encoding idiom - create a "picture" out of the data (what do users see?) ▪ interaction idiom - how do users change what they see? ▪ Design space of the combination of visual encoding and interaction idioms is very large ▪ data and task abstractions help to reduce the number of potential visual encoding and interaction idioms ▪ decision about good or bad matches based on human abilities (visual perception and memory)
8. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 8 March 10, 2022 Word Tree Example
9. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 9 March 10, 2022 Algorithm ▪ Implementation of visual encoding and interaction idioms ▪ can design different algorithms to realise the same idiom ▪ Various factors might impact the choice of a specific algorithm ▪ computational complexity (performance) ▪ memory usage ▪ level of match with visual encoding idiom ▪ Separate algorithm design (computational issues) from idiom design (human perception issues)
10. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 10 March 10, 2022 Threats to Validity ▪ Each design level has their own threats to validity ▪ wrong problem, wrong abstraction, wrong idiom or wrong algorithm
11. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 11 March 10, 2022 Validation Approaches
12. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 12 March 10, 2022 Validation Approaches … ▪ Can perform an immediate or downstream validation ▪ downstream dependencies add to the difficulty of validation - e.g. poor algorithm design may have a negative effect when validating an interaction technique ▪ use of mock-ups for early downstream evaluation ▪ Mismatches ▪ mismatch between the level at which the benefit is claimed and the chosen validation methodology - e.g. benefit of new visual encoding idiom cannot be validated by measuring the performance of the algorithm used downstream ▪ carefully select the subset of validation methods matching the levels of design where contributions are claimed
13. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 13 March 10, 2022 Domain Validation ▪ A field study can help to validate that we are going to address real user needs ▪ observe people in real-world settings ▪ semi-structured interviews (e.g. contextual inquiry) ▪ Downstream validation can for example investigate a solution's adoption rate by the target audience ▪ see what target users do (without brining them into a lab)
14. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 14 March 10, 2022 Abstraction Validation ▪ Identified task abstraction and data abstraction might not solve the target audience's problems ▪ Downstream validation includes testing the solution with members of the target audience ▪ anecdotal (qualitative) feedback whether the tool is useful ▪ field study to observe and document how the target audience uses the tool in their real-world workflow - observe changes in behaviour rather than documenting existing work practices
15. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 15 March 10, 2022 Idiom Validation ▪ Justify the design of the idiom with respect to known perceptual and cognitive principles ▪ Heuristic evaluation or expert reviews may be used to ensure that no known guidelines are violated ▪ Downstream validation ▪ controlled experiments in a lab setting (lab study) - controlled experiments for testing the performance of specific idioms - measure time and errors for given tasks ▪ presentation and qualitative discussion of results - show images or videos of the solution to the target audience ▪ quantitative measurement of resulting visualisations (quality metrics) such as the number of edge crossings for node-link graph ▪ usability studies
16. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 16 March 10, 2022 Algorithm Validation ▪ Analyse computational complexity of algorithms ▪ number of items in the dataset, number of pixels, … ▪ Downstream validation ▪ execution time ▪ memory consumption ▪ scalability ▪ Correctness of algorithm ▪ does implementation meet the idiom specification ▪ Standard benchmarks might help to compare algorithms
17. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 17 March 10, 2022 MatrixExplorer
18. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 18 March 10, 2022 MatrixExplorer Validation Methods
19. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 19 March 10, 2022 Genealogical Graphs
20. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 20 March 10, 2022 Genealogical Graphs Validation Methods
21. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 21 March 10, 2022 Flow Maps Migration from California Top ten states that sent migrants to California (green) and to New York (blue)
22. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 22 March 10, 2022 Flow Maps Validation Methods
23. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 23 March 10, 2022 LiveRAC
24. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 24 March 10, 2022 LiveRAC Validation Methods
25. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 26 March 10, 2022 Sizing the Horizon Validation Methods
26. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 27 March 10, 2022 Exercise 4 ▪ Analysis and Validation
27. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 28 March 10, 2022 Further Reading ▪ This lecture is mainly based on the book Visualization Analysis & Design ▪ chapter 4 - Analysis: Four Levels for Validation
28. Beat Signer - Department of Computer Science - bsigner@vub.ac.be 29 March 10, 2022 References ▪ Visualization Analysis & Design, Tamara Munzner, Taylor & Francis Inc, (Har/Psc edition), May, November 2014, ISBN-13: 978-1466508910
29. 2 December 2005 Next Lecture Data Presentation