Monthly Update: May, 2024
VisML series, AI for Vis, FILWD reading clubs, summer goals, readings and thoughts ...
With the end of the semester, things have been quite hectic here. Keeping up with the publishing pace (one post a week) has been more challenging than usual. Yet, I have managed to publish three posts as planned, even though I had a hard time publishing on Tue, my preferred publishing day.
Looking back it seems like this month has been all about AI/ML! I had not planned to be so focused on this topic, but somehow, this is how the stars aligned.
On a different note, I just finished teaching my PhD-level Information Visualization course and I hope I will be able to report on my experience in a future post. The last time I taught a similar course was in Fall 2022, so it was definitely a special experience for me.
Series on Visualization for Machine Learning
In April I published two new posts of my series on Visualization for Machine Learning.
The first one focused on defining who needs visualization for ML and the second one on what kind of data we have available from models and what we can do with it.
Finalizing these posts required a lot of thinking but I am happy with the final result. There is definitely way more work to do in creating systematic descriptions of these concepts but I think it’s a good initial approximation.
The series will explore specific techniques much more deeply starting with the next post. I will start with visualization techniques to visualize model output or input-output relationships and then move on to visualizing model explanations and model internals. My goal is to go beyond providing a laundry list of techniques and really contribute a way to think about visualizing ML that goes beyond individual methods. I am not sure if I will be able to make it but I will try.
Another issue is that describing individual techniques in my posts may turn out to be too hard or not very expressive, so I am considering recording a few videos to describe these techniques in a more natural way.
As I develop the series I am also considering whether this could eventually become a book. I think this is too preliminary now but it would be really cool to turn this effort into something more tangible and permanent! Can you please let me know by using the poll below?
Please add a comment below if you want to elaborate on your response. I’d love to hear your opinion.
AI for Visualization
Somehow I also managed to write something about the role of AI in visualization.
I had already posted something related previously when I reported on the panel we organized at Northeastern University on a similar topic. That meeting got me thinking more deeply about these ideas and I felt I had something to say. There is absolutely no doubt that we will see a lot of AI in visualization and data science in the next few years. The question is whether these tools will produce desirable advancements or create more problems. In my first posts I focused more on the positive, trying to outline some interesting ways in which AI could contribute to visualization. In my mind, this is a preliminary list I’d like to perfect and propose to my AI colleagues as a challenge. I am so used to talking about what visualization can do for others that I can easily forget to ask what others can do for visualization.
The article gained quite a lot of traction, which was unexpected for me. This confirms that there is a big “appetite” for anything AI these days. Somehow I am not as curious and excited as most people are about AI, but it’s been fun to write this article. I am thinking of writing a second one more focused on what AI cannot do (yet) for data visualization. I started drafting one, but I could not fully convince myself which of the things I had listed AI would not be able to do soon. You will probably see more thoughts from me in this area over the summer.
FILWD Reading Club
Last week our second FILWD reading club took place. This time, we focused on Visualization Literacy for Children. The group read these two related articles and we spent one hour discussing them:
Basak et al. “Visualization Literacy at Elementary School.” In Proc. of the ACM Conference on Human Factors in Computing Systems, CHI 2017.
Chevalier et al. “Observations and Reflections on Visualization Literacy in Elementary School.” IEEE Computer Graphics and Applications 38 (3): 21–29.
I will write a whole report on it soon but here I want to mention that I have been tweaking the format and things are getting better. This time, we reduced the number of papers from three to two, and the timing was better. We also had two participants volunteering to be the timekeeper and the notetaker, which worked really well. There are more adjustments to make (most probably going down to one single paper), but I am reasonably happy with the format already.
The reading club is still a bit in stealth-ish mode because I want to perfect the format before expanding it to more people. In any case, if you are reading this and want to participate in the next one, follow this link and add yourself to the mailing list.
Summer goals
Now that the semester ended I have time to do more work for FILWD. My top priority is to record more video lectures. You may recall from previous posts that I have been meaning to complete my library of videos for the RhetVis course. Now, I finally have time to make this happen. Incidentally, the video lectures will also serve me well next semester when I teach the in-person version of the course to design students at Northeastern University again. The VisML series is also a top priority. Now that I have started it is essential for me to keep going and complete it in a reasonable amount of time. I also want to experiment more with short video interviews. There are a few that I am organizing so you will probably see some soon. Of course, there will be way more posts sprinkled here and there on topics that I cannot even anticipate. I like to write about new things that cross my mind here.
Readings and thoughts …
Last month I finished reading Slow Productivity and I am still thinking about it. It is a super timely book, and it matches very well the kind of work philosophy I have built for myself in the last few years. The ideas around slow productivity naturally led me to think more about how we do things in academia and my frustration with the many distortions I see around me. I have been meaning to write more about what I think are dangerous trends in academia, but I have some trepidations because I want this newsletter to focus on positive messages; there are enough people complaining about what is wrong in the world. However, there is a chance I will post one or two articles sometime soon if I manage to find the right constructive tone.
On a more uplifting note, reading Slow Productivity led me to re-discover the book Rework and the associated podcast, also called Rework. I have been on a full listening binge, and I found so many gems. Jason Fried and David Heinemeier Hansson are incredible. They are the good kind of contrarian I really like. They are original, fun, and yet very pragmatic. And they just don’t give a shit about anybody. I love their attitude! They made me reflect once more about how I organize my work, both the academic and the work I do here. There is simply no shortcut. What matters is doing the work, not talking about the work. Meetings are toxic. Social media won’t save you if you don’t have the content. Etc.
That’s all for now. I am looking forward to the work planned for this month!
—
Your comments, suggestions, and encouragement are always welcome. Even a single line means a lot to me. It makes me feel there are actual people reading this beyond the numbers I see in my Substack dashboard.
If you enjoy reading FILWD and my associated activities, please help me spread the word by sharing my posts on your preferred social media sites, talking about it with your friends and colleagues, and by encouraging people to sign up.
Thanks!!! 🙏
This was the perfect read this morning. Yes, write the book!
I feel that you should write a book on the topic of VizML. It's a problem space that hasn't really been explored to the degree of finding clarity and signal for a wide range of audience/stakeholders. I am really curious how you would tackle visualizing the "black box" in terms of techniques and tools to make it transparent to build trust and understanding. Unless i am unaware of what others are doing in this space, this lines up with a blue ocean strategy