Martin Saveski @msaveski Twitter profile

Pinned Tweet

Martin Saveski

2 years

I’m thrilled to share that I will be joining the @UW_iSchool as an Assistant Professor in the Fall of 2023. I’m looking forward to getting to know my new colleagues and engaging with the stellar social data science community at @UW .

52

7

473

Last Seen Profiles

@BomberosMad

@ian_matveev

@sexckuro

@HlwKshkhh54144

@earthlykisssed

@Tandho_

@ehon_jp

@hypebot

@Prestisa_id

@AdamScehfterw

@AlreadyDeadTPNE

@Niklemore

@ii2xr_

@seattlestoner16

@Alberto37126025

@uf0_m

@Nkityomassa

@tongtong0401

@wsmly4479120641

@Nils_Wollny

@selo77077

@haopakhan

@_tinyplanet_

@vstfld

@nkasi0bim

@UCA_ES

@6inchlena

@prv_se

@banana_child

@Dani19Twink

@cb1crypto

@Tra_Briggs

@TomMaycock3

@stw_pdg

@NCIAorg

@BobbleHatt

Martin Saveski

@msaveski

2 years

[Please RT] I’m recruiting PhD students to start Fall’23 at @UW ! I’m looking for students passionate about using computational methods to study how social platforms can be reimagined to enable better conversations, bridge political divides, and reduce the spread of misinfo. >>

9

263

562

Martin Saveski

@msaveski

7 months

[Please RT] I’m recruiting PhD students to work with me at @UW ! I’m looking for students passionate about using computational methods to study how social platforms can be reimagined to enable better conversations, bridge political divides, and reduce the spread of misinfo. >>>

4

167

295

Martin Saveski

@msaveski

7 years

Excited to present our design for Detecting Network Effects in Randomized Experiments @CODEConference (Session: Network Structure)

1

18

70

Martin Saveski

@msaveski

6 months

If you are at #NeurIPS2023 stop by poster # 925 in the morning session. I will be presenting our work on “Counterfactual Evaluation of Peer-Review Assignment Policies”, done with my extraordinary collaborators Steven Jecmen, Nihar Shah, and @jugander .

0

3

37

Martin Saveski

@msaveski

2 years

I’m especially looking for students with a background and/or interest in social network analysis, machine learning, causal inference, or natural language processing. DM/email me with any questions. @UW @UW_iSchool

1

3

32

Martin Saveski

@msaveski

7 months

I’m particularly interested in the role of algorithms in social platforms. I’m especially looking for students with a background and/or interest in Social Networks, Machine Learning, Causal Inference, or NLP. DM/email me with any questions. @UW / @UW_iSchool

1

3

27

Martin Saveski

@msaveski

3 years

Thanks for the award @icwsm ! I'm happy to see that my reviews were useful and honored to contribute to the community.

ICWSM

@icwsm

3 years

Congratulations to all of the winners of our ICWSM-21 Best Reviewer Award 🎆🌠! Your contributions are essential to building our fantastic program! #ICWSM2021

4

7

87

1

0

23

Martin Saveski

@msaveski

2 years

En route to @ICWSM ! Come check out our two papers on political polarization online (“Session 1: Polarization 1”.) First in-person conference since the pre-times. I’m very excited to catch up with old friends and meet new ones. Come say hi!

0

2

20

Martin Saveski

@msaveski

6 years

If you are at #TheWebConf ( #www2018 ) today stop by the *Web and Society II* session to hear @ngillani1 talk about our paper “Me, My Echo Chamber, and I: Introspection on Social Media Polarization” (). We are truly honored to get an honorable mention for it.

2

4

19

Martin Saveski

@msaveski

2 years

Being on the job market was stressful (ppl were right) but also (surprisingly) fun. It was a pleasure to interact with so many amazing people along the way. Huge thanks to my mentors and collaborators for their support, especially @jugander , @dkroy , @ladamic , and @deaneckles .

1

0

18

Martin Saveski

@msaveski

6 years

Very excited to see my officemate's ( @CrashTheMod3 ) excellent work on the cover of Science! False news spread farther, faster, deeper, and more broadly than the truth news.

Science Magazine

@ScienceMagazine

6 years

How false news spreads, gut microbes that help and harm, and the long-term effects of climate change on fisheries in our newest issue:

12

104

159

0

4

17

Martin Saveski

@msaveski

8 years

Demis Hassabis (DeepMind) @ MIT talking about General AI and AlphaGo.

2

3

17

Martin Saveski

@msaveski

3 years

Today I learned that the easiest way to embed fonts in a PDF is just to open it MacOS Preview and save it. That’s it. Most conferences require that the camera-ready PDFs have all fonts embedded, but most plotting libraries (including ggplot) don't embed the fonts by default.

1

2

17

Martin Saveski

@msaveski

6 years

Lately, I have been using Mac’s text-to-speech feature (cmd + Q) to “proof-read” my writing. It’s a great way to find semantic mistakes that are syntactically correct.

1

0

17

Martin Saveski

@msaveski

4 years

My first attempt to make a Calder mobile.

0

15

Martin Saveski

@msaveski

7 years

@seanjtaylor I went through this exercise recently while putting together a reading list for my general exam. Here is what I got:

0

1

16

Martin Saveski

@msaveski

6 years

No more LaTeXit! It’s nice to be excited about updating software once in a while :)

0

3

16

Martin Saveski

@msaveski

5 years

I find it hard to trust code that doesn’t follow the 80 column width rule (it doesn’t have to be 80, any width is fine as long as it's consistent). Not sure if it’s a good signal, but I think that it does say something about the coder’s attitude.

2

0

16

Martin Saveski

@msaveski

6 years

Interesting blog post on balanced graph partitioning by @googleresearch : . I wish they compared their method with the restreaming techniques in -- reLDG always worked best for me; plus it's so easy to implement.

1

4

15

Martin Saveski

@msaveski

7 years

NPR coverage of FlipFeed: our experiment on bursting social media bubbles.

NPR

@NPR

7 years

Analysts have blamed technology for creating an online echo chamber, but can technology also fix it?

4

35

44

0

4

11

Martin Saveski

@msaveski

6 years

Yesterday this function (just 1 extra line) helped me reduce the computation time of an I/O intensive job from 8 to 1.5 hours! The cachetools & cacheout packages provide similar decorators for other caching strategies.

Jake VanderPlas

@jakevdp

6 years

One of my favorite Python 3 builtins is functools.lru_cache(): with a simple decorator, repeated function calls become O(1) table lookups.

22

393

1K

1

11

Martin Saveski

@msaveski

6 years

I really like this study by @lajello & @nicola_barbieri : it's an excellent example of using observational causal inference to study the effects of recommender systems. ()

Claudia Wagner

@clauwa

6 years

Really interesting AB test on tumblr to investigate the potential danger of recommender systems by @lajello #netsci2018

0

19

1

10

Martin Saveski

@msaveski

7 years

Really enjoyed this talk by Granovetter:

2013 Everett M. Rogers Award Colloquium

Stanford Professor of Sociology Mark Granovetter is the recipient of the 2013 Everett M. Rogers Award.Cited over 24,000 times, Granovetter's 1973 paper "The ...

www.youtube.com

1

0

10

Martin Saveski

@msaveski

3 years

@informor I can confirm this as someone who was on the other side, as an author, recently. I have never received such kind and constructive reviews before. It boosted my motivation to take the feedback into account and ultimately substantially improved the paper.

0

10

Martin Saveski

@msaveski

5 years

For me, this was the highlight of @CODEConference this year: . I love that story!

2019 MIT Conference on Digital Experimentation (CODE@MIT): Fireside...

Promise and Limits of Observational Causal Inference. Featuring Eytan Bakshy, Facebook. Guido Imbens, Stanford. Eleanor Murray, Boston University. Dean Eckle...

www.youtube.com

0

9

Martin Saveski

@msaveski

7 years

I love this amazingly intuitive explanation of the friendship paradox

1

4

9

Martin Saveski

@msaveski

7 years

One of my favorite papers in the last few years -- such a clever idea; now officially out, I still prefer the long arxiv version though.

Dean Eckles

@deaneckles

7 years

How can you test for spillovers from a treatment in a network? Final version of our paper now available in JASA @Susan_Athey

1

77

178

1

0

8

Martin Saveski

@msaveski

9 years

There is a rumor that someone defended his Phd. @CrashTheMod3 @dkroy @sinanaral @socialmachines http://t.co/DeG8kysgPV

0

3

8

Martin Saveski

@msaveski

7 years

It was great to have @tim_cook visit our lab today!

Tim Cook

@tim_cook

7 years

So impressed by @MIT students & faculty who are finding new ways to tackle the world’s biggest challenges. Thanks for sharing your work!

103

589

4K

1

8

Martin Saveski

@msaveski

6 years

@ryanjgallag @gvrkiran had a great tutorial at ICWSM'17 and published a comprehensive list of references. The tutorial website seems to be down, but I saved his references google doc:

Related work for the tutorial

Related work for the tutorial TO READ https://www.nature.com/nclimate/journal/v5/n8/full/nclimate2666.html (great references for echo chambers) Definitions Leon Festinger. A theory of cognitive...

docs.google.com

1

8

Martin Saveski

@msaveski

6 years

@roydanroy I love and hate that my advisor asks me this question every time I bring up internships. My answer is always similar: (1) I like to work with X’s data, (2) I can run large-scale exps I usually can’t in the lab, (3) I will meet new ppl and escape my lab’s idea echo chamber.

1

0

7

Martin Saveski

@msaveski

5 years

Congrats!

Sinan Aral

@sinanaral

5 years

Excited to announce & welcome two amazing young faculty in @johnjhorton & @amaatouq to the IT Group @MITSloan ! They're leaders in their fields & will help us chart the future of IS & CSS research! Fantastic to call you both colleagues! Cc @erikbryn @twmalone @ewzucker @davidautor

7

20

131

0

7

Martin Saveski

@msaveski

7 years

Great reflection on the 'demo-or-die' culture in the Media Lab. Everything is still as Boyd describes.

danah boyd

@zephoria

7 years

I left MIT's Media Lab 15 years ago. I learned an important skill there: "demo-or-die":

3

98

261

0

2

7

Martin Saveski

@msaveski

6 years

Can't wait to read this!

Johan Ugander

@jugander

6 years

New work with @ajwchin & @deaneckles ! Ambitious recent field experiments have been testing the efficacy of targeting strategies in networks. To contrast strategies, "the first step" is to run a full network survey in 30-200 villages/schools. Hard work! /1

1

12

78

1

0

7

Martin Saveski

@msaveski

7 years

Off to Halifax for #kdd2017 . Looking forward to seeing some of you there!

0

6

Martin Saveski

@msaveski

2 years

@manoelribeiro @LorenzoCov ‘s paper is a good example

Detecting Emotional Contagion in Massive Social Networks

Happiness and other emotions spread between people in direct contact, but it is unclear whether massive online social networks also contribute to this spread. Here, we elaborate a novel method for...

journals.plos.org

1

0

6

Martin Saveski

@msaveski

7 years

@CODEConference Papers: &

0

2

6

Martin Saveski

@msaveski

9 years

Why I worry experimental social science is headed in the wrong direction

1

0

3

Martin Saveski

@msaveski

6 years

Really enjoyed reading Newman’s “Mixing patterns in networks” paper (). Every piece of it so elegant and well done. It’s also nice to see what networks researchers cared about back in 2003.

1

0

5

Martin Saveski

@msaveski

3 years

@naz_andalibi I did several before and during my Ph.D., and all were excellent experiences. Three reasons stand out for me: (1) working with great ppl, (2) data access, (3) computing tools/infrastructure. /1

1

0

4

Martin Saveski

@msaveski

2 years

@infoxiao @jmhessel Beautiful! I'm moving to Seattle next year, and I'm excited to experience these long days :). (I'm a bit worried about the short days in the winter, but hey ...)

1

0

4

Martin Saveski

@msaveski

6 years

One of my favorite textbooks!

MIT CSAIL

@MIT_CSAIL

6 years

How you know you’re old: the “Introduction to Algorithms” book was published #otd nearly 30 years ago, in 1990. It’s sold half a million copies & is the world’s most cited CS text w/9K citations:

6

126

309

0

1

4

Martin Saveski

@msaveski

3 years

@deaneckles @mjskay ’s R package ggdist has beautiful geoms for visualizing uncertainty:

0

2

4

Martin Saveski

@msaveski

6 years

@deaneckles @gvrkiran @mitidss Congrats @gvrkiran !

0

4

Martin Saveski

@msaveski

5 years

The art of presenting complex ideas in a simple way. This NYT piece on DNA testing does a great job at that. Analogy and visualization make the ideas so clear.

0

3

Martin Saveski

@msaveski

6 years

@random_walker The key here is the difference between *descriptive norms* (what’s commonly done) and *injunctive norms* (perceptions of what is commonly approved or disapproved within the culture). Injunctive norms can eliminate the negative effects of descriptive norms:

0

4

Martin Saveski

@msaveski

2 years

@hanlinliii @snchancellor Congrats!

0

3

Martin Saveski

@msaveski

4 years

@jeremyzyang @INFORMS @deaneckles @dhillon_p @sinanaral Congrats!

1

0

3

Martin Saveski

@msaveski

7 years

Very cool IV analysis of peer-effects in running by @CNicolaides and @sinanaral

Christos Nicolaides

@CNicolaides

7 years

Social networks push runners to run further and faster than their friends --My new paper with @sinanaral

1

10

61

0

3

Martin Saveski

@msaveski

2 years

@lajello @TahaYasseri @IC2S2 @lau_retti @ITUkbh @DTUtweet @koebenhavns_uni @duncanjwatts Wow! Exciting!

0

3

Martin Saveski

@msaveski

5 years

@estebanmoro It’s a two-sided market, you following the account may significantly increase their likelihood to stay on twitter. I won’t be surprised if they prefer having an new engaged user at the cost of slightly degrading the experience of an already active one.

1

0

3

Martin Saveski

@msaveski

4 years

@deaneckles @jamesheathers This is probably true not just for feature selection but also for hyper-parameter selection as well. I only recently learned about inner and outer CV. (Inner: you do CV to tune your HPs; outer you evaluate the model with the chosen HPs in the inner loop.) \1

1

0

3

Martin Saveski

@msaveski

3 years

@dubeylicious Better yet: ”this seems like a good UROP project”

0

3

Martin Saveski

@msaveski

5 years

@jugander @kaltenburger @Stanford Congrats @kaltenburger !

0

3

Martin Saveski

@msaveski

2 years

@andradaolteanuu igraph + cairo (integrated with igraph) is my go-to for generating static network viz.

0

3

Martin Saveski

@msaveski

5 years

@PaulJWright Thanks! I'm glad you found it useful. Jekyll is a great framework. I tried to design the site so that it's easy to add/edit without touching the html/css code.

1

0

3

Martin Saveski

@msaveski

6 years

@roydanroy Also, flirting with the idea that academia is not the only good choice after grad school imo is very important. Even if a student really wants to go to academia having a plan B lifts off a lot of pressure in the already stressful environment of grad school.

0

3

Martin Saveski

@msaveski

5 years

Excellent combination of machine learning and causal inference!

Ben Letham

@_bletham_

5 years

We just wrote a blog post about using multi-task Bayesian optimization to improve the News Feed ranking model: Paper preprint is also fresh on arxiv ()

0

5

39

0

2

Martin Saveski

@msaveski

6 years

@SohanDsouza @EddyAdelAwad @MITMuseum @MIT @medialab @scalablecoop Nice!

0

2

Martin Saveski

@msaveski

5 years

@KLdivergence Not to justify it but I think it reflects the natural tendency for people to focus on parts of the paper that fit their story and are not necessarily representative of the main takeaway of the paper.

1

0

3

Martin Saveski

@msaveski

8 years

@natematias Here are two great papers: & . One on design, the other on analysis.

0

3

Martin Saveski

@msaveski

8 years

@natematias cool! I would love to hear what are you up to

0

2

Martin Saveski

@msaveski

3 years

@chris_bail @jcccf , @grammarnerd , and @BethanydeGant from FB have published a few excellent papers that answer some of the questions you pose above and I would imagine motivated the later experiments/deployment: &

1

0

3

Martin Saveski

@msaveski

5 years

@jeremyzyang @danielrock @dhillon_p Me too! Lol. And that’s after stairing at the pics for 10s.

0

3

Martin Saveski

@msaveski

4 years

@ryanjgallag Good idea. I also try to do a "tone correction" pass—reading it over and thinking "can I be nicer?" :)

0

1

3

Martin Saveski

@msaveski

7 years

@natematias @FastCompany Cool! Congrats, Nathan!

0

3

Martin Saveski

@msaveski

6 years

@mrfrank5790 @ScienceAdvances @iyadrahwan @ahmad__kareem @lijunsun0 @cesifoti Congrats!

0

3

Martin Saveski

@msaveski

6 years

[Gist] We built a tool that allows Twitter users to explore the politically-active parts of their network and ran an RCT to test how it affects users’ (1) beliefs about their network, (2) political diversity of who they follow, and (3) political alignment of URLs they share.

1

3

Martin Saveski

@msaveski

4 years

I learned many things at @kdd_news 2017. This was one of them.

Canadian Forces in 🇺🇸

@CAFinUS

4 years

This tree from Nova Scotia is now in Boston Common. The Nova Scotians send one every year. Why?

1K

11K

38K

0

3

Martin Saveski

@msaveski

2 years

@tanmit @infoxiao I had a similar problem. Rotating it every 30 secs or so helped.

0

2

Martin Saveski

@msaveski

2 years

@dArsovec @itosheva @andon_ru @MedoDimoski Hahaha—budi spomeni slikata!

1

0

2

Martin Saveski

@msaveski

4 years

@deaneckles @Twitter On the Metrics dropdown menu (top right) you select "Custom Metrics", then you have to select "Total audience reach" (under General).

2

0

2

Martin Saveski

@msaveski

5 years

@johnjhorton It was really fun designing the AI agent! The same techniques were the core of deep blue. I had extra motivation since I did it for a class and we had a tournament at the end of the semester where we played each other's agents. It was limited time/move, so I had to optimize it :)

0

2

Martin Saveski

@msaveski

4 years

@maria_antoniak Ha! I finally switched to VS code for everything except notebooks (still using Jupyter Lab), but this prompts me to give it a shot :)

1

0

2

Martin Saveski

@msaveski

2 years

@srijankedia @NSF @gtcomputing @GTCSE @ICatGT Congrats!

0

2

Martin Saveski

@msaveski

1 year

@JohnWest @ChadSDay @rebeccaballhaus @joe_palazzolo @JamesVGrimaldi @coulterjones @michaelsiconolf Congrats!

1

0

2

Martin Saveski

@msaveski

4 years

@rasbt This is very cool! Thanks for sharing it. I'm curious about your thoughts about the pros and cons of each (maybe you have lecture notes?). Or at least, what's your go-to?

1

0

2

Martin Saveski

@msaveski

5 years

@johnjhorton Very cool! Looks much nicer than my Python2 version from 10 years ago :) []. Though, I have to say that I still feel proud of my search algorithm and evaluation function.

1

0

2

Martin Saveski

@msaveski

2 years

@FaidraMonachou @YaleSOM @HarvardCMSA @skominers Congrats!

0

2

Martin Saveski

@msaveski

7 years

@Noriega_Campero presenting their work on The Wisdom of the Network (cc @amaatouq )

0

1

2

Martin Saveski

@msaveski

8 years

Sad news! Gonna miss @kimonolabs

2

0

2

Martin Saveski

@msaveski

3 years

@daveholtz @sinanaral @deaneckles @johnjhorton @ssuri @AndreyFradkin Congrats! 🥳

1

0

2

Martin Saveski

@msaveski

5 years

@bkeegan Gerber and green's field experiments is my favorite intro CI book. No observational CI ofc, but I think it's better to start with field experiments anyway.

0

1

2

Martin Saveski

@msaveski

2 years

@FrancescoBonchi @TheWebConf Congrats, Francesco!

0

2

Martin Saveski

@msaveski

4 years

@deaneckles @jamesheathers I get really worried about reproducibly when I read all these deep NN papers which only do train/val/test splits, no CV. \2

1

0

2

Martin Saveski

@msaveski

7 years

Talks about the rejection of the Strength of the Weak Ties paper, summarizes the history of SNA & the impact of social nets on movements.

0

1

Martin Saveski

@msaveski

5 years

@jugander @ajwchin @Stanford Congrats, @ajwchin !

0

2

Martin Saveski

@msaveski

5 years

@danielequercia @EdmondAwad I wanted to make an intro but I thought you guys already met. @EdmondAwad you should definitely talk to @danielequercia !

1

0

2

Martin Saveski

@msaveski

6 months

@elenadata @Zara_ftm @ab_asudeh Congratulations, @Zara_ftm !

0

2

Martin Saveski

@msaveski

2 years

@tinaeliassi @KhouryCollege @NUnetsi @Northeastern Congratulations, Tina!

0

2

Martin Saveski

@msaveski

2 years

@raphaelschaad Indeed, but not too short :)

Causal Effects of Brevity on Style and Success in Social Media | Proceedings of the ACM on Human-...

In online communities, where billions of people strive to propagate their messages, understanding how wording affects success is of primary importance. In this work, we are interested in one partic...

dl.acm.org

1

0

2

Martin Saveski

@msaveski

5 years

@bkeegan @AOC very cool! does she tweet more often after the election?

1

0

2

Martin Saveski

@msaveski

4 years

@Leesplez @reimthyme @Prolific Publishing the anonymization code is a great idea!

0

2

Martin Saveski

@msaveski

3 years

@deaneckles Not face-to-face, but we did the Twitter version of this a few years ago:

Me, My Echo Chamber, and I | Proceedings of the 2018 World Wide Web Conference

dl.acm.org

0

2

Martin Saveski

@msaveski

3 years

@naz_andalibi Also, psychologically, it was great to see that people do excellent research outside of academia and consider that as a possible career path. /2

0

2

Martin Saveski

@msaveski

3 years

@daveholtz @sinanaral @deaneckles @johnjhorton @MITSloanPhD Congrats!

1

0

2

Martin Saveski

@msaveski

3 years