Week in Vis 9 Mon, Oct 28-Fri, Nov 1

Last week, we thought about interaction and implementation. I was doing this in Vancouver at the VIS conference. I was also seeing lots of other visualization work, which will probably work its way into class.

This week, we’ll talk about the other “how” of visualization: how people see, and what this means for visualization. A day of lecture and a few readings are not going to turn us into Perceptual Scientists or Psychologists (unless you already are) – so we’ll just be getting a little taste of how the human visual system works (or at least the current theories), and try to think about what this means for us as visualization designers. There’s a bit more reading this week than the past few weeks, but most people find it interesting.

There is a chance I might swap the ICE and lecture (to give me more time to get the new things I learned about perception into the lecture), so the ICE may be on Monday not Wednesday. Don’t be surprised if you show up Monday and we do an ICE rather than a lecture.

Design Challenge 2 has its next phase – a “rough draft” that shows us that you really are working on it, and not saving the assignment to the last minute. I think I called it “signs of life” in the assignment description.

Readings for the Week

The main readings are the Ware chapters, since it’s a good introduction to the basics of perception, and its impact on design. Chapter 6 of Cairo is useful because it considers “higher level” perceptual issues. I also include Cairo Chapter 5 (as optional) because it’s redundant with Ware, but it’s fun to see his (less scientific) take on it. And look at Chris Healy’s web page to get a sense of pre-attentive effects.

I also want you to look at the Healy and Enns paper / resources. It is sufficient to look at the web survey (since it has the cool demos).

  1. Visual Queries (Chapter 1 of Visual Thinking for Design) (Ware-1-VisualQueries.pdf 2.5mb)
  2. What We Can Easily See (Chapter 2 of Visual Thinking for Design) (Ware-2-EasilySee.pdf 2.1mb)
  3. Structuring Two Dimensional Space (Chapter 3 of Visual Thinking for Design) (Ware-3-StructuringSpace.pdf 2.6mb)
  4. Visualizing for the Mind (Chapter 6 of The Functional Art) (theFunctionalArtCh6.pdf 8.1mb)
  5. Look at the pre-attention demos and pictures in the old version of Chris Healey’s web survey of perceptual principles for vis. The paper (optional, below) is much better in terms of explaining things – but it’s too much to require as reading.

Perception: Optional

Perceptual science is a whole field, so we’re just touching the surface. Even just the beginnings of what is relevant to visualization. It’s hard for me not to require these…

  • The Eye and Visual Brain (Chapter 5 of The Functional Art) (theFunctionalArtCh5.pdf 5.4mb) Optional – Cairo’s take on it. More based on his experience as a designer.
  • Healey, C. G., & Enns, J. T. (2012). Attention and Visual Memory in Visualization and Computer Graphics. IEEE Transactions on Visualization and Computer Graphics, 18(7), 1170–1188. (pdf) (doi)

This is a good survey of basic perception stuff that is useful for vis. In this past, this was required reading.
Warning: this survey is a little dense, but it gets the concepts across with examples. Don’t worry about the theory so much. Get a sense of what the visual system does (through the figures, and the descriptions of the phenomena), and skip over the theories of how it does it (unless you’re interested).
There is an older, online version as Chris Healy’s web survey which has lots of cool pre-attention demos. But the text in the paper is much better, and the paper includes more things.

  • Franconeri, S. L. (2013). The Nature and Status of Visual Resources. In D. Reisberg (Ed.), The Oxford Handbook of Cognitive Psychology (pp. 1–16). Oxford University Press. (pdf) (doi)

    This is a survey, similar to Healey and Enns above, but written more from the psychology side. The first part, where he characterizes the various kinds of limitations on our visual system is something I’ve found really valuable. The latter parts, where he discusses some of the current theories for why these limitations happen is interesting (to me), but less directly relevant to visualization (since it is mainly trying to explain limits that we need to work around). I think these explanations may lead to new ideas for visualization – but its less direct of a path.

  • Albers, D., Correll, M., Gleicher, M., & Franconeri, S. (2014). Ensemble Processing of Color and Shape: Beyond Mean Judgments. Journal of Vision, 14(10), 1056–1056. (paper page) (doi)

    We (Steve, myself, and some of our students) have written a survey paper on some other things the visual system can do (and why it can matter for vis). We call it “visual aggregation” and in psychology they call it “ensemble encoding.” It might be useful to skim through for the pictures and diagrams. I will talk about this stuff (at least the work that we did) in class.

DC2 Task List

by Mike Gleicher on October 18, 2019

A list of tasks (compiled from what people have turned in) is posted on canvas.

The grouping (and combining of similar tasks) is Aditya’s judgment (I think it’s a good organization).

This list is meant to inspire your thinking for the later phases of DC2. You don’t have to choose tasks from this list. There are plenty of good choices.

However, beware: there are some that you cannot do with the data we’ve provided!

Week in Vis 8 Mon, Oct 21-Fri, Oct 25

There is the option of adding a “fix” for your DC1. We’re deep into DC2 – don’t forget the sketches due this week!

The topic is interaction, which is interesting in a number of ways. On Monday’s lecture, we’ll look at some ways to think about interaction. On Wednesday, we’ll talk about implementing it and look at examples.

And if you’re wondering, I’ll be learning about new stuff in Vis as well at the IEEE VIS 2019 conference. Aditya will take care of lectures. He’ll show off some of the things that he’s worked on.

Readings for the Week

The first reading is a survey paper that provides a good way to organize many of the interactions we see in visualization, and provides lots of good examples.

  1. Heer, J., & Shneiderman, B. (2012). Interactive dynamics for visual analysis. Communications of the ACM, 55(4), 45. (pdf) (doi)
  2. Maniplate View (Chapter 11 from Munzner’s Visualization Analysis & Design) (Munzner-11-ManipulateView.pdf 0.5mb)
  3. Facet into Multiple Views (Chapter 12 from Munzner’s Visualization Analysis & Design) (Munzner-12-FacetMultipleViews.pdf 1.0mb)This isn’t specific to interaction, but it fits better here than anywhere else.


I’ll use this paper to frame the discussion in class. It provides a good “why not add interaction” point of view.

  • Lam, H. (2008). A Framework of Interaction Costs in Information Visualization. IEEE Transactions on Visualization and Computer Graphics, 14(6), 1149–1156. (doi). (pdf link to Heidi’s page)

DC1 “Bonus”

by Mike Gleicher on October 17, 2019

In grading the first few assignments, we noticed a common pattern. We wanted to give people an opportunity to make a fix if they need to.

As the assignment says: We look for signs of students making explicit good choices to make what they want the viewer to see easy to see. (you can explain your choices in the documentation)

In class, I was more explicit: I told you that your “story” should be “something that is easy to see in the visualization”, and that the assignment is about making explicit choices to make the story be the thing that is easy to see.

I think I even phrased it in class as “the way to get an A is to make design choices that make it easy to see the thing that you want to be easy to see, and to explain what these design choices are.” (Aditya remembers me discussing this).

When we looked over the first few assignments (warning – it could be that we just saw 5 bad ones and 1 good one – and the other 50 will be good), people seemed to miss the point that the “story” should be something that the visualization makes “easy to see.” Maybe I wasn’t clear enough that you should be explicit to explain this. Very few designs were so good that it was obvious (both what stands out and what choices were made to cause it).

In many of the assignments we looked at (from the small sample), we weren’t sure what should be easy to see – and the caption / description didn’t tell us what we were supposed to see, and the description/rationale didn’t explain what was done to make it easy to see!

So, in the event that it was just me not being clear, I am going to give people a chance to do this more explicitly. If you want, you may add an extra page to your DC1. This is optional. If you do it, it can help your grade (although not as much as if you had done it explicitly in the first place). For a specific design or two (at most two), please tell us:

  1. Which of the 5 of your designs are you describing.
  2. What does this design make easy to see? This should be a fact about the data (the “story” – or the finding of the story). The fact should be specific – not just “we can see variable X and Y” – but something like “we can see the positive correlation” or “we can see the pattern that there are more yelp check ins on weekends.” And it should be something that really is easy to see (look at your visualization!)
  3. Why is it easy to see? What choices did you make that makes it easy to see?

You can provide a description in the type in box on this Canvas Assignment. You must do it on or before Monday, October 21st. This is optional. Do not turn anything else in if you think your assignment is good enough (e.g., you already answered this for all of your designs, or the thing that is easy to see is so obvious that we won’t miss it), or don’t think it will help (you realize that your visualization doesn’t make anything easy to see).

Note: one of the first 5 assignments we looked at did this very well and explicitly. They said “here is the thing we can see clearly in the visualization” and described the design decisions they made that caused it to happen. Definitely an A – not just because they followed the instructions, but also because they were good designs (I believe there is a correlation – if you think about “what am I trying to make easy to see” you will be more likely to make good choices to make that thing easy to see). You may not need to add anything if your assignment is already good. This is optional.

Unfortunately, this will delay grading (we want to give people a chance to do this before grading).

Week in Vis 7 Mon, Oct 14-Fri, Oct 18

This week we’ll consider a different kind of scalability challenge: having too many dimensions. This is a hot topic in machine learning, bioinformatics, … Unfortunately, we won’t get to dig too deeply into the mathematics. But, if you’ve seen dimensionality reduction or embedding in an ML or stats class, this might give you some more basic insights.

We also move on to design Challenge 2, with a first phase due on Wednesday. The Wednesday deadline is pretty tight – we want to take your list of tasks and compile them so that everyone can see the whole list.

In class on Wednesday, I’ll take some time to talk about DC2 – but mainly to answer questions. So make sure you’ve thought about DC2 so you have questions to ask.

Readings for the Week

Last week, we focused on scaling in the number of items. This week, we’ll talk about what to do when we have too many dimensions.

  1. High-Dimensional Visualizations. Georges Grinstein, Marjan Trutschl, Urska Cvek. (semantic scholar) (link1)

    This is an old (Circa 2001) paper that I am not sure was actually published at KDD. However, it is a great gallery of old methods for doing “High-Dimensional” (mid-dimensional by modern standards) visualizations. Most of these ideas did not stand the test of time – but it’s amusing to look through the old gallery to get a sense of what people were trying.

  2. The Beginner’s Guide to Dimensionality Reduction, by By: Matthew Conlen and Fred Hohman. An Idyll interactive workbook.

    This is a very basic demonstration of the basic concepts of dimensionality reduction. It doesn’t say much about the “real” algorithms, but you should get a rough idea if you haven’t already.

  3. How to Use T-SNE Effectively

    I wanted to give you a good foundation on dimensionality reduction. This isn’t it. But… it will make you appreciate why you need to be careful with dimensionality reduction (especially fancy kinds of it).

Cancel office hours on Wednesday, October 9th

by Mike Gleicher on October 7, 2019

I need to cancel office hours on Wednesday, October 9th.

I will be at class.

If you need to speak to me outside of class and cannot wait until the 16th, please send me email me.

Design Challenge 2 is Posted!

by Mike Gleicher on October 7, 2019

You can look at Design Challenge 2. The first phase is due next week.

The Week in Vis 06 (Mon, Oct 7 – Fri, Oct 11): Scale

by Mike Gleicher on October 3, 2019

Week in Vis 6 Mon, Oct 7-Fri, Oct 11

This week we wrap up Design Challenge 1 with the peer reviews. They are due on Friday (since we assigned them on Friday), but you might want to get them done. DC2 is coming soon. (it should be posted this week)

The topic this week is scale – how do we deal with problems where we have “too much stuff” (that’s a reference to a song, but it doesn’t matter). We all talk about “big data” – but for vis, how do we deal with things that are big? Big may mean something different (for me, “big data” means “data that is big enough that it becomes challenging”). The ICE is designed to get you to think about strategies for dealing with large data and should set up.

Readings for the Week

These 3 things are required. The Munzner chapters are fairly short, and Alper’s paper will give you a good way to think about scalability more generally.

  1. Reduce Items and Dimensions (Chapter 13 from Munzner’s Visualization Analysis & Design) (Munzner-13-Reduce.pdf 0.4mb)
  2. Embed: Focus+Context (Chapter 14 from Munzner’s Visualization Analysis & Design) (Munzner-14-Embed.pdf 0.5mb)
  3. Sarikaya, Gleicher and Szafir. Design Factors for Summary Visualization in Visual Analytics. (web) – This is a survey of different ways of doing summarization that appear in the visualization literature. There is a lot about how the survey was conducted, but the main thing for class is to see the different categories of summarization and how they interact.

On Discussion Grading

by Mike Gleicher on September 30, 2019

A few notes on the grading of online discussions (both “Online Discussions” and Seek and Finds) that come up:

  1. A piece of general advice for dealing with reviews (this comes from decades of having research papers rejected): put youself in the mindset of “what did I do in my writing that caused the reviewer to … (miss the thing I wanted them to see, come to the wrong conclusion, not appreciate what I wrote, …)”. Learn from it: think “what could I have done to make it easier for the reviewer/grader to get my point.”
  2. There is noise in the process. This is why we average over a large number of observations (well part of the reason, we also want to account for people traveling, etc.).
  3. Putting #1 and #2 together: we need to grade a lot of these quickly, so if you want to be sure you get full credit for your work, make it easy for the grader to see that your posting satisfies the requirements. See Aditya’s advice. It is not enough that you have a “complete answer” – it must be easy for the grader to see that you have a right answer. (they have to go through a lot of things quickly)
  4. Hopefully, you will get a sense of what a good answer is from seeing others in the discussion. In general, we prefer that you do not edit your initial posting after you make it. If you feel like you’ve made a big mistake (for example, seeing other answers you realize you misinterpreted the question), put in an improved answer as a response to your post (and label it as such).
  5. There is an element of subjectivity to grading. That’s part of #2.
  6. An A (90) is the highest grade you can get. If you ask us to reconsider your A, you are telling us that you do not think you deserve it.
  7. Remember that we have to deal with a lot of these. Some small mistakes will happen, and we cannot always give lots of feedback. (see #2). We simply don’t have the resources to give everyone detailed feedback on every assignment, and small mistakes will happen. We try to give enough feedback that you can learn from it, and let the statistics work out the noise.
  8. We don’t like to argue with students about grades. If you say “I think my answer is complete/correct”, we are likely to say “yes, but I can also see how the grader would miss this, work on making your response clearer next time” – but we may say “I agree with the grader that this answer is not as good as the others that received higher grades, I think the grader was too generous.”

Tips for Online Discussions

by Aditya Barve on September 28, 2019

I saw great conversations on Canvas during Online Discussions 1 and 2. That’s a fantastic start!

A few simple changes could significantly improve future discussions.

Addressing the prompt

Make sure your response addresses most, hopefully all, of the prompt’s points. It is easy to get sucked into one idea and forget to address the rest of the prompt.

A welcoming initial post

Structure your response in paragraphs. Headings are a bonus. Headings could be the question you are answering or the theme of your paragraph(s). The principles of visualization can also be applied to format text effectively. Also, avoid run-on sentences.

If your post is difficult to understand, your peers are less likely to engage you in a discussion.

A meaningful comment

Some of you only say nice things. Others just point out possible flaws. A little more balance and variety would benefit everyone. For instance, you may want to comment when a post:

  • caused a eureka moment (share your revelation)
  • confused you (ask for clarification)
  • has hidden assumptions (talk about them)
  • has assumptions you disagree with (discuss whether they are justified)
  • reminded you of a cool example (explain it, perhaps include a photo)