Arvind Satyanarayan
@arvindsatya1
Followers
5,623
Following
1,679
Media
285
Statuses
7,557
Assistant Professor @MIT_CSAIL @mitvis . Data visualization @vega_vis , ML interpretability, cognitively convivial interaction. He/him. @arvind @vis .social.
Boston, MA
Joined October 2007
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
FEMA
• 1256813 Tweets
Mancuso
• 114620 Tweets
Best Record
• 97460 Tweets
#turkishwomenareindanger
• 93601 Tweets
Grammy
• 84650 Tweets
Ülke
• 74137 Tweets
Pogba
• 72660 Tweets
Uyuşturucu
• 65700 Tweets
Joker 2
• 61080 Tweets
Valencia
• 35815 Tweets
#ikbaluzuner
• 35528 Tweets
Dolly
• 28984 Tweets
Marseille
• 28333 Tweets
#yasaistiyoruz
• 26782 Tweets
Flint
• 25268 Tweets
Jamie Dimon
• 23942 Tweets
Semih Çelik
• 23924 Tweets
Celtics
• 19662 Tweets
Neymar
• 19082 Tweets
Angers
• 16236 Tweets
Laura Bonaparte
• 15565 Tweets
Leeds
• 14778 Tweets
#hassasiçerik
• 12201 Tweets
Greenwood
• 11225 Tweets
İdam
• 10651 Tweets
Pinned Tweet
Last month, I had a lot of fun giving a lightning talk about some new ideas I'm stewing on regarding generative AI and human agency.
It touches on only the first piece of a longer argument we unpack in a recent white paper.
Summary 🧵👇
1
11
53
This tweet crosses the line. Critiquing the substance of a piece of research is one thing. But, it is absolutely unacceptable to target a PhD student in this way.
Shame.
Senior scientist work hard & write hundreds of papers to earn recognition.
Social media & preprint servers are turning a model successful for hundreds of years up-side-down.
Compare data reported by Google Scholar before dismissing giants like John Ioannidis. Shame on MIT.
217
149
541
13
107
1K
Super excited to publish "The Building Blocks of Interpretability," a
@distillpub
article where we explore rich user interfaces for understanding neural networks.
5
369
822
Great analogy on the current state of conducting/publishing research from
@Protohedgehog
#ieeevis
#openscience
7
130
266
Especially in this time of high uncertainty, I’m glad to have a source of personal anxiety resolved: after 13 years in the US, I am finally holding a green card in my hands!
11
4
247
One of my favourite parts of teaching Interactive Data Visualization (
#mit6894
) is seeing the incredible final projects students produce. We usually celebrate them with a poster session, but we're going online this year: . Here're some highlights:
4
47
218
Crystal is an incredibly skillful researcher as seen in the adept and ambitious way she combined computational and ethnographic approaches in her CHI 2021 paper. She is an enormous asset to every community she participates in. I and
@mitvis
continue to learn a lot from her.
0
17
208
After 4 months on the road (and an extra 7lbs!), I'm thrilled to announce I'll be joining
@MIT_CSAIL
as an Asst. Prof. in July 2018!
33
12
171
17
11
153
Honored to be in such great company, and very grateful to all the folks at
@mitvis
whose work this fellowship is really recognizing!
We have today announced the names of the 2024 Sloan Research Fellows! Congratulations to these 126 outstanding early-career researchers:
6
40
248
33
6
141
VisText has been a *ton* of work, and 2 yrs of solid effort. So it's exciting to (finally!) be able to talk about it, and it's gratifying to see it featured on
@MIT
's homepage.
Lead author,
@bennyjtang
, has a great thread w/details below 👇
And I wanted to offer a few thoughts
Chart captioning is hard, both for humans & AI.
Today, we’re introducing VisText: a benchmark dataset of 12k+ visually-diverse charts w/ rich captions for automatic captioning (w/
@angie_boggust
@arvindsatya1
)
📄:
💻:
#ACL2023NLP
3
25
126
2
28
131
Bittersweet to wrap up my time
@GoogleBrain
. I'm grateful to have worked with
@ch402
,
@shancarter
,
@ludwigschubert
, and
@enjalot
. Thanks to their generosity and patience, I learned a whole bunch that will undoubtedly shape the research my lab will do.
Onwards to
@MIT_CSAIL
!
5
2
127
I feel unspeakable rage reading these anonymous messages to a survivor of sexual harassment.
The survivors represent the very best of us, and of computer science. We are not deserving of their bravery, and we will be forever in their debt for dragging us to a more just future.
0
10
120
Very grateful to receive this recognition. Working with
@mitvis
folks has been such a source of joy (especially over the pandemic) and I’m really humbled they thought to nominate me for this honor.
Fifteen MIT faculty honored as “Committed to Caring” for 2021-23: Honor recognizes professors who went the extra mile advising during the pandemic’s disruptions.
1
16
106
3
4
109
As an HCI researcher, I'm heartened by
@drfeifei
&
@mijordan3
advocating for more human-centric perspectives in ML research. I'm excited to help make
@distillpub
the premier venue for publishing these ideas and anticipate benefits to articulating them within an interactive medium
We're really excited that
@arvindsatya1
has accepted a position as a Distill editor! Arvind, who is about to start as a professor at MIT CSAIL, will have a special focus on articles at the intersection of ML & HCI.
3
14
122
0
12
107
📢 Come join us at
@mitvis
!! We're looking to recruit 1–2 PhD students to work with us on accessible data representations, interpretable/explainable AI, or malleable interfaces w/LLMs.
Apply to MIT EECS by Dec 15:
(lists a lot of helpful resources!).
6
31
106
After more than a year of work, super proud to release Vega-Lite 2 (and Vega 3). Excited to build on this foundation moving forward.
Introducing Vega-Lite 2.0: A grammar of *interactive* graphics, powered by Vega 3's reactive dataflow engine.
1
65
175
3
27
102
I was delighted to be a keynote speaker at
@hildaworkshop
(one of my favourite academic venues because it bridges various communities). I talked about work we've been doing at
@mitvis
and with collaborators towards _effective_ interaction with data vis.
1
10
99
Excited to kickoff our inaugural Boston VIS PhD Summer School!
Thanks to my co-organizers
@FILWD
and
@artistJaneAdams
(who gets extra major props for designing our logo!)
6
7
97
I'm VERY excited about this postdoc position joint between us (
@mitvis
) and
@MITanthropology
! We're want to use linguistic/semiotic anthropology to connect performances of MAGIC 🪄✨and VISUALIZATION 📊📈
Come work with us!!
4
26
84
I've posted my NSF CAREER proposal here: . Good luck to everyone submitting!!
1
7
79
I'm very excited about this type of article. It feels like formally recognizing that the hard work of peer review is a research contribution as well: we saw both the original authors and respondents invest considerable effort replicating results and running new experiments.
A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features' - Six comments from the community and responses from the original authors.
6
200
615
1
12
75
Just got to
#chi2022
! Please come say hi — especially if we don’t know each other!! Students: I’ll be happy to treat you to a beverage of your choice because it is HUMID out!
3
0
74
After 15 months away, everything is exactly as we’d left it… almost like it had been encased in amber. I’m struck by how much joy I’m getting from being back in the building, and what a shot of motivation and energy it’s given me that even a week of vacation did not!
1
0
70
The steady march of time through university ID cards!
5
0
68
Hey
#ieeevis
folks, are you teaching a vis course that needs to go online? Rather than individually recording/livestreaming lectures, I wonder if we can produce better content together? E.g., lecture pod/videocasts that are conversations between 1-2 faculty? Who'd be interested?
8
15
64
Honored our work on interactive Vega-Lite won Best Paper. Excited to present at InfoVis '16.
7
24
63
Feature visualization show neural nets build an understanding layer-by-layer. From
@ch402
,
@zzznah
&
@ludwigschubert
1
19
64
👏👏👏
@Birdbassador
’s fantastic, provocative, and thoughtful
#Vis4DH
keynote is up as a Medium article:
#ieeevis
2
22
60
Yes, I have trouble understanding how it is in the "public safety" to keep the identities of perpetrators anonymous after a decision to ban has been made. It imposes a burden on victims to be fortunate enough to be connected to whisper networks.
Michigan's behavior has been so bad here that it paints ACM in a good light, but do remember that ACM did not reveal who the prof was, and asked victims not to reveal. So that ACM can just stand on the side and watch him harass more women. Pathetic.
3
4
61
2
7
59
Wrapping up our second BostonVIS PhD Summer School!
HUGE THANK YOU to student organizers
@WoottonDylan
and Sydney Purdue!!!
And thank you to our faculty facilitators
@FILWD
,
@laneharrison
,
@Birdbassador
,
@vizstudylady
, and Jieqiong Zhao!
0
7
59
HUGE CONGRATS to
@bennyjtang
and
@angie_boggust
for receiving an
#ACL2023NLP
Outstanding Paper Award for VisText!!! 👏🎉🔥
Chart captioning is hard, both for humans & AI.
Today, we’re introducing VisText: a benchmark dataset of 12k+ visually-diverse charts w/ rich captions for automatic captioning (w/
@angie_boggust
@arvindsatya1
)
📄:
💻:
#ACL2023NLP
3
25
126
2
4
58
A fun little Colab notebook that
@ludwigschubert
,
@ch402
and I worked on to produce feature inversion caricatures. They're helpful for understanding the abstractions a model learns without depending on basis directions to align with human concepts.
2
16
59
Crystal is an incredibly skillful researcher as seen in the adept and ambitious way she combined computational and ethnographic approaches in her CHI 2021 paper. She is an enormous asset to every community she participates in. I and
@mitvis
continue to learn a lot from her.
0
17
208
0
0
57
Slides from my
#openvisconf
talk on Reactive Vega and Interactive Vega-Lite available here:
"Reactive Building Blocks: Interactive Visualizations with Vega" with
@arvindsatya1
#openvisconf
0
3
6
0
24
55
Hooray!
@domoritz
,
@halden_lin
,
@jeffrey_heer
, et al. win InfoVis Best Paper for Draco: Formalizing Visualization Design Knowledge as Constraints.
#ieeevis
3
8
55
Great talk! I'd a similar track record at the start: my first 1st-author paper (Ellipsis) was rejected 4 times and then just scraped in, my second 1st-author paper (Lyra) was brutally rejected at CHI and then squeaked in. Rocky years when I wasn't sure I was cut out for research.
Coping with Rejection in Academia: A Three-Pronged Approach.
15
83
297
0
0
54
After three years, it’s really great to be at a conference with and in the same room as some of my datavis faves!
1
0
53
.
@enjalot
shows how t-sne can be applied more generally by pointing to
@ch402
’s excellent post, which has greatly influenced my thinking on vis + ML.
#openvisconf
0
12
51
This is one of the most thorough VIS/HCI systems papers I’ve ever read! Rich and detailed motivation, design rationale and (most importantly!) tradeoffs. Incredible figures.
Wow wow wow. It’s set a standard for the rest of us to aspire to!
Fantastic work
@hyeok__kim
et al!!
Our CHI 2022 paper introduces “Cicero,” a declarative grammar that expresses transformations for RV in an expressive, flexible, and reusable way. (2/10)
1
7
53
3
4
51
2
0
51
3
1
51
Last week, at the first (of hopefully many!)
@VisualDataSci
at
#kdd2021
, I was grateful to share some thoughts on systems-building for ML interpretability occurring in HCI/VIS. And how, at
@mitvis
, we're slowly shifting from individual systems to toolkits.
0
6
48
Thank you
@dalbersszafir
, it’s an honor to share this recognition with you!!
#ieeevis
2
7
85
5
0
45
Congrats to
@seankandel
, Andreas Paepcke,
@joe_hellerstein
, and
@jeffrey_heer
for their 10yr Test of Time!! 🎉📈
📃Paper:
(Ps that’s three-in-a-row ToTs for
@jeffrey_heer
! Phew, what an inspiration!)
#ieeevis
1
6
45
Shocking despicable behavior. Bravo to the graduate students for sharing their stories publicly, and to the student journalists for reporting this out.
“My experience was a part of a pattern of behavior”
The Daily’s first issue of a two part investigation into allegations of sexual misconduct against CSE professor Walter Lasecki.
4
96
197
1
8
44
One of my favorite parts of the article is when we begin to formalize this insight as a design space. Low-hanging future work: enumerate all interpretability interfaces and conduct comparative evaluations to understand what they are/aren't good for.
1
9
44
Huge thx to folks who've helped over the yrs, esp.
@jeffrey_heer
who has been the consummate advisor. Excited to carry the
@uwdata
torch fwd
After 4 months on the road (and an extra 7lbs!), I'm thrilled to announce I'll be joining
@MIT_CSAIL
as an Asst. Prof. in July 2018!
33
12
171
2
4
43
I'll be defending my PhD thesis at 2pm today! Follow along w/the livestream: and slides:
4
5
43
HUGE CONGRATS to
@ohnobackspace
!! 🎉🎉
@cuinfoscience
is very lucky to have you, and I can’t wait to see what you and your students work on together! 📈 🎇
hi! i will be presenting this work at
#chi2024
next week.
i'm also pleased to share that i will be joining the faculty at
@cuinfoscience
in 2025, and intend to recruit students soon.
my dms are open if you'd like to chat at the conference. thanks!
35
11
169
0
2
41
I really appreciate the forcefulness of President Reif’s email. “ICE is unable to offer the most basic answers about how its policy will be interpreted or implemented. [...] Unspoken, but unmistakable, is one more question: Am I welcome? At MIT, the answer, unequivocally, is yes”
0
7
40
I'm pretty excited about this new tenure-track position between our College of Computing and Dept. of Architecture that's connecting data, design, and visualization!
More details here:
0
5
39
Wendy Mackay has been such an incredible mentor, a huge source of inspiration, and so formative to my career that I am DELIGHTED she is being recognized with the Lifetime Research Award—
@sigchi
’s highest honor.
CONGRATS Wendy!!! So very well deserved!! 🎉🎉
1
1
39
Congrats
@mitchellgordon
!! VERY excited to welcome you as a colleague
@MITEECS
@MIT_CSAIL
next year!
Such stellar work done by our very own
@mitchellgordon
. Mitchell is such a wonderful researcher, lab mate, and friend!
@MITEECS
is super lucky to have him next year as he starts his role as an Assistant Professor!!! Super well deserved...
1
4
45
1
2
38
Warm fuzzies as we wrap our first Boston Vis PhD Summer School!
Thank you students for participating actively, and generously sharing anecdotes + advice!
Thank you faculty facilitators
@laneharrison
,
@vizstudylady
,
@michelle_borkin
for keeping it real!
Excited to kickoff our inaugural Boston VIS PhD Summer School!
Thanks to my co-organizers
@FILWD
and
@artistJaneAdams
(who gets extra major props for designing our logo!)
6
7
97
1
4
38
In addition to the research contribution, which
@alanlundgard
describes in this thread, one of the things I'm proud of is that we converted the gnarly
#ieeevis
PDF into an accessible HTML page. Plan to share our workflow soon, which is based off LaTeXML:
Descriptions, captions, and alt text can help make visualizations more accessible. But what should they say, or not say? My new paper with
@arvindsatya1
contributes a model for evaluating textual descriptions of charts.
Accessible HTML Paper:
#a11y
/Thread
5
70
246
2
4
37
Our key insight is that we should be *combining* interpretability techniques (not studying them in isolation). E.g., with feature vis + attribution, we can compare two classifications (floppy ears apparently distinguish types of dog!).
1
15
37
We’ve talked quite explicitly about taking things easier this summer at
@mitvis
. I’ve strongly encouraged everyone to take some serious time off. If that means fewer
#chi2022
submissions, that’s a-ok. There’s always another deadline.
The case for a genuine (manager-facilitated) summer slowdown:
2
41
194
0
1
37
Every yearish, we host a summer camp for new+existing VIS junior faculty, to build community & peer mentor.
We've just started recruiting the next 2-yr cohort. If you're interested, plz fill out this form:
More info:
#ieeevis
1
13
37
.
@jeffrey_heer
on Stu Card’s advice: “toolkits not only codify knowledge, but thru their design instantiate a theory of doing.”
#ieeevis
2
11
36
Congrats Jonathan!!! Such well-deserved recognition of the care, thoughtfulness, and scope of your research vision! It's been an honor working alongside you these past few years, and a genuine pleasure watching you grow into your researcher identity!🥲📈💥
0
4
36
1
0
36
For each of our interface ideas, we also provide colab notebooks so that you can try it out with your own input images. E.g., reproduces the class comparison UI. HackerNoon has a good overview of how to run them with Google's GPUs:
1
13
35
Hooray!! Such a well-deserved recognition! This paper was one of the first things I read as a PhD student that got me excited about working on data visualization (and working with
@jeffrey_heer
specifically!).
Eddie Segel & I's work on Narrative Visualization won an
#ieeevis
InfoVis 10 Year Test-of-Time Award. We are humbled and honored, and also shell-shocked by the speed of time's arrow (and just about everything else these days). Here is our decade-old paper:
21
33
276
0
0
35
After 500+ commits, proud to announce the release of Vega 2.0 featuring streaming data and declarative interaction.
http://t.co/DjLSCZVUur
1
18
32
It was a joy to critically reflect on Lyra, Data Illustrator, and Charticulator with
@bongshin
,
@donghaoren
,
@jeffrey_heer
,
@johntstasko
,
@jr_thomp
,
@mattbrehmer
, and
@zcliu
#ieeevis
2
11
35
I like this framing about data fallacies from
@datassist
: fallacies occur when the data answers a different question from the one that was asked.
#openvisconf
1
10
33
We're increasingly hearing a consistent story emerge from our community about why they've adopted Vega/Vega-Lite. Robin does a great job of articulating many of these points in his Medium post.
I just published “Why I’m backing Vega-Lite as our default tool for data visualisation”
1
10
58
0
20
34
A suit and tie didn’t feel right for me either. But I still wanted to project “professor” since I certainly didn’t _feel_ like a professor yet. And I wanted something that made me feel good/confident.
So, here’s what I went with (all credit goes to my partner for picking these!)
2/ When I went on the market (2000), I decided I would not wear a suit or tie. I wore khaki/tan-ish pants. Not super-formal but neat (no cargo pockets), dark-colored button-down long-sleeve shirts, and a jacket ("sports coat"?). That's it. ↵
2
0
13
3
0
32
.
@shancarter
draws a fantastic analogy between hidden layers and artistic movements. Cubism, surrealism, etc. were, in a sense, experimenting at a specific level of visual abstraction.
#openvisconf
0
9
32
Great insight from
@3blue1brown
. The way Math is currently taught (equations -> examples) is like starting a building from the top floor down. Nothing to ground the abstract concepts in. But once you see elegance of the eqn, hard to unroll your understanding back.
#SpreadingFacts
2
2
30
Very proud of how we’ve been able to expand the
#ieeevis
Inclusivity & Diversity program this year. Shout out to my I&D co-chairs
@michelle_borkin
and Kelly Gaither
3
4
31
Woo hoo!!!! Congratulations Professor Lee!!
I'm SO glad that you'll continue to be at MIT — we're very luck to have you!
1
1
31
.
@aboutaaron
summarizing a fantastic W. E. B. Du Bois quote: “visualization can facilitate liberation”
#openvisconf
1
12
30
Flapjax really influenced our (
@uwdata
) work with designing declarative interaction design primitives in
@vega_vis
It's worth reading
@ArjunGuha
's "acceptance speech" of our Flapjax Test-of-Time Award, because it's basically half of PLMW in a few paragraphs. Grad students: take courage. Your instincts are right.
9
52
200
2
5
31
HUGE CONGRATS to
@joshmpollock
et al. on Bluefish's alpha release!! 🎉🍾 🎆
I've been SO excited about Bluefish for a very long time!! (Especially because of the pipeline of other equally-exciting things its allowing us to incubate!)
Excited to share Bluefish, a new JS library for making diagrams!
Bluefish lets you code diagrams like they're UIs: using declarative components and reactivity.
Check it out:
Quick backstory 🧵
11
69
331
0
4
29
Today, I am canceling my meetings and pausing active research in support of
#ShutDownAcademia
, and have encouraged members of
@mitvis
to do the same.
1
2
28
So excited for this major milestone!! A HUGE thank you to
@domoritz
for leading the charge on this effort.
The Vega Projects now have a charter, steering committee, documented governance, and easier ways for people to become maintainers at .
1
6
49
0
1
28
A recent preprint in this vein that I’ve really appreciated is from Ashley Suh and friends (incl
@sirrice
) on a hypothesis grammar. Such a richly generative contribution!
I'm increasingly interested in higher-level support for analysis tasks. Grammars are a powerful way for (both people and machines!) to say "make this chart". But most grammars are agnostic to the larger task and intent. Could we (should we?) bridge that gap?
7
3
28
2
3
28
Why is visualization a bad neighbor?
@Birdbassador
notes that by binning data we lose underlying nuance and humanity in the data. Subsequent visual forms then make it easy to forget that nuance exists at all.
#ieeevis
#Vis4DH
2
4
28
Really cool to see
@mathisonian
and
@fredhohman
share analytics and feedback to their
#ieeevis
#visxai
explainable “The Beginner’s Guide to Dimensionality Reduction”
0
5
28
Really love how my MacBook Pro's butttterfly keyboard is actinng up tthe day before tthe
#EuroVis
deadlinnne. Itt's the cherry onnn ttop of ttthe sundae!
1
0
28
Super love this point from
@accidental_PhD
: how we represent our data (tabular or network) is an abstraction that shapes what we think we can do with it, or how we can work with it.
#ieeevis
3
9
27
Congrats
@crystaljjlee
!!! 🎉🎉🎉
Congrats to
@crystaljjlee
for being awarded a Social Data Dissertation Fellowship for her work on “Visualizing a Pandemic: Epistemological Conflicts over Covid-19 Data on Social Media”
1
3
44
2
1
27
.
@kennelliott
's curation of human perception studies remains one of my favourite go to lists on this topic.
#CJ2016
0
14
26
“Please be as forgiving as you yourself would want your
reviewers to be unto your own submissions.” Such a great letter from the papers chairs, with advice that holds true beyond the current chaotic period.
1
4
27
@ShriramKMurthi
I find classic HCI theory on direct manipulation really helpful for teasing apart the nuance here. Namely, textual and non-textual specification can both have the same semantic distance but differing articulatory distances.
3
5
26
Great slide from
@mathisonian
talking about how
@ParametricPress
has straddled the line between research/practice, and how the two lenses have helped reveal unique opportunities and approaches.
#ieeevis
1
8
25