My excuse to attend this year (besides the close proximity) was that another class project was accepted as a poster. Juliette Pardue, Mridul Sen (@mridulish), and Christos Tsolakis took a previous semester's project, FluNet Vis, and generalized it. WorldVis (2-pager, PDF poster, live demo) allows users to load and visualize datasets of annual world data with a choropleth map and line charts. It also includes a scented widget in the form of a histogram showing the percentage of countries with reported data for each year in the dataset.
Before I get to the actual conference, I'd like to give kudos to whomever picked out the conference badges. I loved having a pen (or two) always handy.
Monday was "workshop and associated events" day. If you're registered for the full conference, then you're able to attend any of these pre-main conference events (instead of having to pick and register for just one). This is nice, but results in lots of conflict in determining which interesting session to attend. Thankfully, the community is full of live tweeters (#ieeevis), so I was able to follow along with some of the sessions I missed. It was a unique experience to be at a conference that appealed not only to my interest in visualization, but also to my interests in digital humanities and computer networking.
I was able to attend parts of 3 events:
- VizSec (IEEE Symposium on Visualization for Cyber Security) - #VizSec
- Vis4DH (Workshop on Visualization for the Digital Humanities) - #Vis4DH
- BELIV (Beyond Time and Errors: Novel Evaluation Methods for Visualization) Workshop - #BELIV
The VizSec keynote, "The State of (Viz) Security", was given by Jay Jacobs (@jayjacobs), Senior Data Scientist at BitSight, co-author of Data-Driven Security, and host of the Data-Driven Security podcast. He shared some of his perspectives as Lead Data Analyst on multiple Data Breach Investigations Reports. His data-driven approach focused on analyzing security breaches to help decision makers (those in the board room) better protect their organizations against future attacks. Rather than detecting a single breach, the goal is to determine how analysis can help them shore up their security in general. He spoke about how configuration (TLS, certificates, etc.) can be a major problem and that having a P2P share on the network indicates the potential for botnet activity.
In addition, he talked about vis intelligence and how CDFs and confidence intervals are often lost on the general public.
Now @jayjacobs arguing that we need visual literacy in support of security — optimal charts often hard to interpret. #VizSec #ieeevis— Lane Harrison (@laneharrison) October 24, 2016
He also mentioned current techniques in displaying IT risk and how some organizations allow for manual override of the analysis.
Fair criticism from @jayjacobs on traditional risk registers & heat maps, & the impact of manually overriding model results #ieeevis #vizsec pic.twitter.com/6JEBeNFuE9— Savannah Fitzwater (@Atomic_Fitz) October 24, 2016
And then during question time, a book recommendation: How to Measure Anything in Cybersecurity Risk
In addition to the keynote, I attended sessions on Case Studies and Visualizing Large Scale Threats. Here are notes from a few of the presentations.Great ? from audience about how to improve heatmaps. @jayjacobs recommends "How to Measure Anything in Cybersecurity Risk" #ieeevis #vizsec— Savannah Fitzwater (@Atomic_Fitz) October 24, 2016
"CyberPetri at CDX 2016: Real-time Network Situation Awareness", by Dustin Arendt, Dan Best, Russ Burtner and Celeste Lyn Paul, presented analysis of data gathered from the 2016 Cyber Defense Exercise (CDX).
"Uncovering Periodic Network Signals of Cyber Attacks", by Ngoc Anh Huynh, Wee Keong Ng, Alex Ulmer and Jörn Kohlhammer, looked at analyzing network traces of malware and provided a good example of evaluation using a small simulated environment and real network traces.
"Bigfoot: A Geo-based Visualization Methodology for Detecting BGP Threats", by Meenakshi Syamkumar, Ramakrishnan Durairajan and Paul Barford, brought me back to my networking days with a primer on BGP.
"Understanding the Context of Network Traffic Alerts" (video), by Bram Cappers and Jarke J. van Wijk, used machine learning on PCAP traces and built upon their 2015 VizSec paper "SNAPS: Semantic Network traffic Analysis through Projection and Selection" (video).
"An alert does not stand on its own." Bram Cappers argues for context-understanding tools for network traffic alerts. #ieeevis #vizsec— Steven R. Gomez (@steveg_cs) October 24, 2016
Fantastic demo from Cappers and van Wijk's work on ConTA at #VizSec. If you missed it, a video is up at https://t.co/e6ewqGAD1t #IEEEVIS— Sophie Engle (@sjengle) October 24, 2016
DJ Wrisley (@djwrisley) put together a great Storify with tweets from Vis4DH.
Here's a question we also ask in my main research area of digital preservation:Super excited to attend Visualization for DH workshop #VIS4DH #ieeevis & looking forward to sharing when I get back @uglibrary @DHatGuelph! pic.twitter.com/wwmv04cxkU— Lucia Costanzo (@luciacostanzo) October 24, 2016
A theme throughout the sessions I attended was the tension between the humanities and the technical ("interpretation vs. observation", "rhetoric vs. objective"). Speakers advocated for technical researchers to attend digital humanities conferences, like DH 2016, to help bridge the gap and get to know others in the area.
There was also a discussion of close reading vs. distant reading.
AJ Bradley at #ieeevis #VIS4DH: "I don't care how many nouns are in this poem" - on the absurd focus on deriving insights from vis— matt brehmer (@mattbrehmer) October 24, 2016
Distant reading, analyzing the structure of a work, is relatively easy to visualize (frequency of words, parts of speech, character appearance), but close reading is about interpretation and is harder to fit into a visualization. But the discussion did bring up the promise of distant reading as a way to navigate to close reading.
I made the point to attend the presentation of the 2016 BELIV Impact Award so that I could hear Ben Shneiderman (@benbendc) speak. He and his long-time collaborator, Catherine Plaisant, were presented the award for their 2006 paper, "Strategies for Evaluating Information Visualization Tools: Multidimensional In-depth Long-term Case Studies".
I also attended the "Reflections" paper session, which consisted of position papers from Jessica Hullman (@JessicaHullman), Michael Sedlmair, and Robert Kosara (@eagereyes). Jessica Hullman's paper focused on evaluations of uncertainty visualizations, and Michael Sedlmair presented seven scenarios (with examples) for design study contributions:
- propose a novel technique
- reflect on methods
- illustrate design guidelines
- transfer to other problems
- improve understanding of a VIS sub-area
- address a problem that your readers care about
- strong and convincing evaluation
I propose to establish the new Robert Kosara's Mantra "How Do We Know That?" #ieeevis.— Enrico Bertini (@FILWD) October 24, 2016
Tuesday, October 25
The VIS keynote was given by Ricardo Hausmann (@ricardo_hausman), Director at the Center for International Development & Professor of the Practice of Economic Development, Kennedy School of Government, Harvard University. He gave an excellent talk and shared his work on the Atlas of Economic Complexity and his ideas on how technology has played a large role in the wealth gap between rich and poor nations.
When Adam Smith wrote his book, wealth gap between richest and poorest country was 4:1, today it is 256:1. Ricardo Hausmann at #IEEEVIS— Ming the Merciless (@eagereyes) October 25, 2016
Hausmann argues that the difference between rich and poor is due to technology. #IEEEVIS— Ming the Merciless (@eagereyes) October 25, 2016
Technology isn’t just tools and encoded knowledge, but critically the tacit knowledge that’s deep in people’s brains, nowhere else #IEEEVIS— Ming the Merciless (@eagereyes) October 25, 2016
Tools as embodied knowledge; codes as codified knowledge; knowhow as tacit knowledge @ricardo_hausman #ieeevis— Tamara Munzner (@tamaramunzner) October 25, 2016
Online version of @ricardo_hausman product space visualization https://t.co/nQBkcJraV7 (United States, 2014) @HarvardCID #dataviz #ieeevis pic.twitter.com/1LBiHJwpzq— Romain Vuillemot (@romsson) October 25, 2016
After the keynote, in the InfoVis session, Papers Co-Chairs Niklas Elmqvist (@NElmqvist), Bongshin Lee (@bongshin), and Kwan-Liu Ma described a bit of the reviewing process and revealed even more details in a blog post. I especially liked the feedback and statistics (including distribution of scores and review length) that were provided to reviewers (though I didn't get a picture of the slide). I hope to incorporate something like that in the next conference I have a hand in running.When the Treemaps father @benbendc ask questions to the Treemaps fan and everyday user @ricardo_hausman #ieeevis pic.twitter.com/UXnPHqrlpJ— Romain Vuillemot (@romsson) October 25, 2016
I attended parts of both InfoVis and VAST paper sessions. There was a ton of interesting work presented in both. Here are notes from a few of the presentations.
"Visualization by Demonstration: An Interaction Paradigm for Visual Data Exploration" (website with demo video), by Bahador Saket (@bahador10), Hannah Kim, Eli T. Brown, and Alex Endert, presented a new interface for allowing relatively novice users to manipulate their data. Items start out as a random scatterplot, but users can rearrange the points into bar charts, true scatterplots, add confidence intervals, etc. just by manipulating the graph into the idea of what it should look like.
"Vega-Lite: A Grammar of Interactive Graphics" (video), by Arvind Satyanarayan (@arvindsatya1), Dominik Moritz (@domoritz), Kanit Wongsuphasawat (@kanitw), and Jeffrey Heer (@jeffrey_heer), won the InfoVis Best Paper Award. This work presents a high-level visualization grammar for building rapid prototypes of common visualization types, using JSON syntax. Vega-Lite can be compiled into Vega specifications, and Vega itself is an extension to the popular D3 library. Vega-Lite came out of the Voyager project, which was presented at InfoVis 2015. The authors mentioned that this work has already been extended - Altair is a Python API for Vega-Lite. One of the key features of Vega-Lite is the ability to create multiple linked views of the data. The current release only supports a single view, but the authors hope to have multi-view support available by the end of the year. I'm excited to have my students try out Vega-Lite next semester.
Best Paper at InfoVis goes to Vega Lite by @arvindsatya1, @domoritz, @kanitw, and @jeffrey_heer. Congrats! https://t.co/LXxkPSeBSM #IEEEVIS— Ming the Merciless (@eagereyes) October 25, 2016
"HindSight: Encouraging Exploration through Direct Encoding of Personal Interaction History", by Mi Feng, Cheng Deng, Evan M. Peck, and Lane Harrison (@laneharrison), allows users to explore visualizations based on their own history in interacting with the visualization. The tool is also described in a blog post and with demo examples.
"PowerSet: A Comprehensive Visualization of Set Intersections" (video), by Bilal Alsallakh (@bilalalsallakh) and Liu Ren, described a new method for visualizing set data (typically shown in Venn or Euler diagrams) in a rectangle format (similar to a treemap).Introducing HindSight, a design technique for interaction history - #ieeevis— Lane Harrison (@laneharrison) October 24, 2016
Demos: https://t.co/hQZFgrmfnz pic.twitter.com/VFMPBvxsFf
On Tuesday night, I attended the VisLies meetup, which focused on highlighting poor visualizations that had somehow made it into popular media. This website will be a great resource for my class next semester. I plan to ask each student to pick one of these and explain what went wrong.
Although I was only able to attend two days of the conference, I saw lots of great work that I plan to bring back into the classroom to share with my students.
(In addition to this brief overview, check out Robert Kosara's daily commentaries (Sunday/Monday, Tuesday, Wednesday/Thursday, Thursday/Friday) at https://eagereyes.org.)