Loading...
 
Print

Education Workshop 2010

Welcome to the Education Workshop 2010

August 30, 2010

University of Maryland

Conference Center

Rm. 2100


Point of contact: Kris Cook


agenda

Draft Taxonomy

Syllabi

Education Resources


This page contains the notes taken during the workshop. We encourage the participants to edit this page both during and after the workshop so that it accurately captures the workshop content.


Observations, Common Threads and Surprises


  • Students don't like Tableau, but instructors saw it as being easy to use
  • Importance of case studies and studies, but need repositories for data
  • Individual instructors are singing their own songs - need to come together to create something that could be used at an undergraduate level
  • How does this differ from multi-variate statistics?
    • If you work from top-down, rather than bottom up, you see the connection to task.
    • Visual analytics has cognition associated with it as well
    • Visual analytics requires dynamic interactivity
    • Fundamental: making informed decisions. Visual analytics is the means to an end.
  • Theme: how do we slice and dice visual analytics in an educational setting? The challenge of interdisciplinary work in the context of siloed contexts.
  • We're doing OK within the visual analytics / infovis community. But how do we grow into the technical communities that surround us?
  • Need to engage decision sciences and cognitive sciences more than we currently are.
  • One course isn't enough.
  • Data scientist vs. programmers. There would be different courses in visual analytics for users vs. visual analytics for developers. Visual analytics for users is a pre-requisite for the developer course.
  • Three-person teams: content/data specialist, tool developer, and communications specialists.
  • Challenge of creating a curriculum and scholars that are truly interdisciplinary. How to educate students (not just machine learning students or infovis students) so that they address the aspects and are the mix?
  • Should we be teaching to the teachers? Professors don't necessarily feel prepared to give a full visual analytics course.
  • There is no textbook for visual analytics. A longterm goal could be to create a textbook.
    • GPU for example has an out-of-the-box course that you can teach.
    • Could a textbook be a living document online that could be free and updated continually?
    • Single-module vs. multi-modules - subdivide it into multiple modules.
  • Software take advantage of new hardware / parallelism offers new ways to explore data.
  • How are we going to evaluate what people have learned via visual analytics course?
  • Nobody's focused on what the job market is like for visual analytics, especially for undergraduate and graduate students.



What kinds of graduates should visual analytics programs produce?

  • People who are unafraid of complex problems and complex data
  • Tool developers - application developers (research or industry). Emphasis on engineering.
  • Theoretical and applied research in visual analytics.
  • Analysts vs. tool developers?
  • Students have to be willing to interact with domain scientists.
  • In the cognition community:
    • Base level of analytic competence: be a tool builder
    • Level 2: Researcher
    • Level 3: Publishing in the cogsci community
  • Students who can recognize and present patterns. Have ability to analyze a lot of different complex data problems.
  • Analysts who can make informed decisions on complex, messy data.
  • There is a continuum from analyst to developer - data scientist. Analysts who can also develop the tools they need, for example.
  • Challenge to the group: we're aiming too low. A goal should be to make visual analytics to accessible to people in other fields.
  • Grad students in education not currently using visual thinking, but this would be an area for outreach.

Knowledge that graduates should have

  • Mass communication
  • Understanding language of other technical domains
  • Statistics
  • Reflective problem-solving
  • Understanding of strengths and weaknesses of approaches
  • User advocacy
  • Interaction methodology
  • Algorithms - nP hard identification
  • Understand limitations to quantitative methodologies
  • Ability to deal with massive datasets
  • Knowledge of distributed systems and modern computing
  • Design
  • Artificial intelligence and expert systems


Skills graduates should have

  • Communication skills
  • Mash-up experts
  • Visually literate
  • Computational skills - know how to program
  • Visualization skills
  • Data manipulation and analysis skills
  • Teambuilding skills

Discussion

How to distribute these skills in a degree or certificate program? Will be easiest for computer science skills?

Does a person need to have all of these, or just some subset?

Needs to be a course that shows how each of these elements are unified.

There needs to be a core set of technologies that is necessary for all, but then you can specialize from there.

We've evolved from a visual analytics course to a curriculum to a faculty of visual analytics. However, if visual analytics becomes a part of all other fields, then it becomes absorbed and does not stand alone.

Analogy to the data mining community's evolution. The group came up with the life cycle through all of those separate portions. Focus on the full life cycle first, then teach the focus on the particular specialties in those course.

There has to be an ability to go beyond the limitations of existing tools.

It would be worthwhile to take the visual analytics work flow and build an extended taxonomy that maps to the multiple fields.

What are the core things that are held in common?

There is a lexicon of visualization that will not change. Assume a few constants.

Formal dogma: data to information to knowledge to insight. Abstract away all the issues.

What would be the minimum for visual analytics course?

  • Basic knowledge of data analysis technique
  • Basic knowledge of visualizations
  • Appreciation for users' problems and tasks
  • Appreciation of cognitive issues (bias, etc)
  • Know the current state of the art in tools



Draft outline of a visual analytics survey course syllabus

Two options:
  • Course 1A: broad survey. Prerequisite is problem solving abilities. Focused on definition of this course.

    • 1 semester, 3 hrs/week = 45 hours.
    • Targeting Advanced undergrad, beginning grad
    • Prereq - Problem solving ability

    • 1.5 hrs - Introduction
    • 1.5 hrs - Initial case study
    • 5 hrs – Cognition, Perception, Human Visual (color, C. Ware, Memory)
    • 6 hrs - User tasks, Sensemaking, User centered design (tasks, reqs, HCI, lol)
    • 8 hrs - Data analysis techniques, data transformation, clustering, data cleaning★
    • 3 hrs - Analysis Tools
    • 8 hrs - Interactive visualization techniques, visualization techniques, **Interaction, Graph/networking★
    • 3 hrs - Telling stories (Tableau, MayEyes, Jigsaw)
    • 3 hrs on evaluation, communication, design, briefing, storytelling, HCI
    • 6 hrs on case studies
    • Design taught throughout

★can be interchanged


  • Course 1B: programming. Prerequisites include introductory programming (possibly algorithms) and college mathematics (such as calculus)
This would include all of 1A, plus programming assignments.


Draft outline of a visual analytics course sequence

  • 101 - Survey course
  • 200s - Electives
    • 201 - Seminar. Papers; applications in specific domains; multiple options.
    • Computer science
      • 202 Data analysis (M, Stats, DM, KR)
      • 203 Visualization (Interaction)
    • Cognitive science
      • 204 Problem solving (creativity, analytic reasoning)
      • 205 Decision-making
    • either CS or Cog Sci
      • 206 HCI User Evaluation
  • 301 - Capstone: thesis, paper, project, etc. could target others like biology, geography, etc. modified depending on class enrollment.

Framework can be used for graduate or undergraduate curriculum.
If some of these courses could be mapped to existing courses, that would be easier to get implemented in a university.

Issues

  • Need to develop a textbook
    • need to better define which areas need the additional resource
    • collection of papers, chapters in books, and other materials instead of a textbook
    • different contributors write chapters rather than one author
  • How is data visualization different from information visualization
  • How do we keep the labs we've created funded
  • How to get funding - war council on how to maintain survival
  • If you taught a VA course who/what would it focus:
    • Application users
    • Tool developers
    • ...
How did it go with respect to your audience (CS, social sci, cognition)
  • If you taught a VA course what did it include:
    • Visualization
    • Analysis
    • Cognition
    • Perception
    • Writing/Presentation
  • What is the intellectual basis for VA? Is it just another name for multivariate stats or computational stats or data mining?
  • How is VA different from data scientist?
  • Where is the evidence that this approach has enabled us to get new insight not available from other techniques/disciplines?
  • Need to define what careers exist for VA students.
  • 3 fields of graduate
    • 1) Mass communication
    • 2) Data mining/machine language
    • 3) Specify in fields
      • Psychology
      • Biomedicine
      • Language
      • Geography
      • Business
      • Law
  • How do we incorporate ethics into VA courses?
  • Accessibility for users with visual impairments
  • How do we actually make sure that all this turns into real courses given in our institutions?

Actions

  • Complete list of resources for educators section of this site
    • Owner - All
    • George
    • Liz
  • Take Intro Survey course filled in hour by hour
  • Need business model - aplia.org as example
  • Find a venue to get together again (not VisWeek - too busy)
  • Identify the terminology used in VA and map across domains
  • Build a representation of the Visual Analytic process
  • Outreach to journalists
    • Professional societies
    • Technical magazines
    • General Press (WSJ, NYT, CNN, CBS)
  • Outreach to economists:
    • Point out the "marginal benefit" of visual analytics in
      • dealing with unstructured data sets (Tukey/EDA/formulating the questions) as opposed to structured data sets more easily examined with traditional statistical methods
      • dealing with "missing" data e.g. filling in the gaps using real time news feeds to supplement regional economic data
      • producing mashups to put fractured data sets in some comprehensible context for decision making purposes
  • Encourage website to unify community (like KD nuggets)
  • Collect success stories on the site
  • Outreach to education (to develop a feed/seeding system for future visual analysts)
    • Collaborate with Schools of education e.g.:
      • Find the faculty members responsible for teaching "Methods" an upperclass course generally required of education majors before they do student teaching. If you can reach these student teachers with VA skills when they are most impressionable, they will seed the schools they visit with new teaching methods and presentations to senior teachers
      • Find the K-12 organizations that are doing research and creating "visual thinking" materials
      • Reach out to work with CCSSO - Council of Chief State School Officers
      • Ditto National Council for Geographic Education - They are already "into" VA
    • Work with doctoral candidates in school of education to learn/visualize data sets on education http://nces.ed.gov









Page last modified on Wednesday, September 08, 2010

Print