Murat Seçkin Ayhan

Assistant Professor of Computing and Software Systems, UW Bothell; Previously UCL and University of Tübingen

[Murat Seçkin Ayhan @msayhann]

I have moved to the breathtaking Pacific Northwest and joined the ranks of the University of Washington in Bothell.

Here’s to new opportunities, growth, and discovery! 🥂🧿🍀

[Murat Seçkin Ayhan @msayhann]

@prof_kamilov And there is no refund. Sale is final.

[tetsuo @tetsuoai]

I LOVE YOU❤️in C.

[Peyman Milanfar @docmilanfar]

also, addition is a special case of multiplication

[Hannah Lucinda Smith @hannahluci]

In today's @MonocleMag newsletter I make the case for absorbing 'kolay gelsin', the most useful of Turkish phrases, into English.
Happy Monday, and kolay gelsin!
https://t.co/mz3TeZYnkX

[Murat Seçkin Ayhan @msayhann]

Mic dropped!

[Oded Rechavi @OdedRechavi]

Every PI

[Murat Seçkin Ayhan @msayhann]

Journals, authors, universities…
Among all these entities, we need a better credit assignment system. A friend of mine proposed this with his student:

https://t.co/Rb4Dk4T2N9

Unless we stop gamification, we will only chase our tails.

[Oded Rechavi @OdedRechavi]

I think I’ll start posting about the lessons I’m learning as part of this new thing I’ve been doing (my attempt to change the landscape of scientific publishing and consequently how science is done)

One lesson I’ve learned (and also unlearned…) is that it’s very convenient to put all the blame on journals. I’ve done it myself for years. And yes, many of the criticisms are valid. They make way too much money at our expense and are often not very good at distinguishing good science from bad science. Some of them (not all of them! There are good journals too!) bring very little value and can even slow scientific progress. They can be inefficient and biased, and journal names are a very poor substitute for quality.

But the more I work on this, the harder it is for me to believe that journals are the only problem (even specifically when it comes just to publishing science). Universities are equally at fault. And I don’t just mean that we, the scientists doing the reviewing, are part of the problem (which we are, obviously). I mean the institutions we belong to, and the way they make decisions. Hiring, promotions, funding allocation - these processes are often opaque, subjective, and not particularly scientific. They are slow, inefficient, and they rely on journal brands as a shortcut.

I used to think journals were driving this, but it’s obviously more like a loop. Journals could not stay the way they are if universities changed how they evaluate quality, because they would lose much of their justification to exist. But universities do not evaluate science directly, because there is too much of it and not enough experts available and time (or money to pay reviewers). So they rely on journal prestige, while journals rely on institutional reputation. Where you do your science ends up mattering more than what you discover, and this affects publication, which affects funding, which determines whether you can even pursue your ideas.

This can be exploited, of course, but I don’t think institutions (or the responsible faculty/management) behave this way because they are evil or greedy. They do it because evaluating science properly is ridiculously hard and time-consuming, and the system does not reward doing it well.

But the important question is can we change the way our universities work, or is it an impossible task? What I've learned working on this problem is that we can. In addition to engaging with management we can influence the system in other ways. In many cases we don’t need their approval. We are the ones who form the committees. I believe we can break the loop, if we target the mechanism of science evaluation. Journals will keep their power, shortcuts will keep dominating, and the same biases will keep reproducing themselves unless we change how we evaluate science (how we do review). If we can find ways to critically evaluate science at scale, rigorously and transparently, we can change how decisions are made.

[Murat Seçkin Ayhan @msayhann]

@docmilanfar Afiyet olsun.

[Murat Seçkin Ayhan @msayhann]

@ftm_guney Yalan dünya…

[Murat Seçkin Ayhan @msayhann]

@ealpaydin Quicksort has been cited only 1924 times. Not novel. Its impact is limited.

[Murat Seçkin Ayhan @msayhann]

@CSProfKGD German döner provides only a Laplacian approximation to the original.

[Murat Seçkin Ayhan @msayhann]

@andrewgwils Same goes for principles over papers and citations, as well.

[Murat Seçkin Ayhan @msayhann]

@pmddomingos A friend of mine wrote a nice piece on this with his student: https://t.co/Rb4Dk4T2N9

[Murat Seçkin Ayhan @msayhann]

@izmirgibiyiz Yiyecektimiş…

[Murat Seçkin Ayhan @msayhann]

@CSProfKGD A friend of mine wrote this piece together with his student: https://t.co/y4zZeF3UB1

[Murat Seçkin Ayhan @msayhann]

Postdoc’s transformation

[Oded Rechavi @OdedRechavi]

The "Authors Contribution Statement" means absolutely nothing

[Murat Seçkin Ayhan @msayhann]

I still remember my number! 😃

[World of Engineering @engineers_feed]

Are you this old? Ah-oh!

[Murat Seçkin Ayhan @msayhann]

A computer is a serious tool, Michael!

[Retro Recall (☥𝐃𝐁) @RetroMoviesDB]

Kids today won't understand how hard this intro goes.

[Murat Seçkin Ayhan @msayhann]

Legends never die! 🏎️💨💨

[Ferrari News TR @ferrarinewstr]

12 yıl önce bugün: Efsanemiz Michael Schumacher, geçirdiği kayak kazasının ardından komaya girdi.

O günden beri savaşı hala sürüyor ama bu savaşta yalnız değil. Zaferlerin en güzeli en sona saklanır. İnanmaya devam edeceğiz.
#KeepFightingMichael

[Murat Seçkin Ayhan @msayhann]

@CSProfKGD Even Andrew Ng made a quick pass over NNets in ML Class in 2011, before the inception of Coursera. Back then, kernel methods and Bayesian learning were making the headlines. I don’t think they exclude each other though. All of them are interesting and puzzling.