Hacker News new | past | comments | ask | show | jobs | submit login
Bret Victor Redesigns Classic Strogatz Paper (2011) (worrydream.com)
206 points by jgamman on Apr 14, 2018 | hide | past | web | favorite | 50 comments

First, I have to say I love the redesigned paper. It's extremely intuitive and easy to read even for people with very little background. In text form I imagine it would be accessible to ten times fewer people at least.

I don't think all papers should be done this way, but it would be fantastic if more were. Pop-science explanations of papers are almost always too simplistic and almost nobody's going to read plain study texts unless they have to.

How to explain (somewhat) difficult concepts and ideas has been a sort of hobby of mine for a while, everything from contributing to simple.wikipedia to Reddit comments explaining new studies.

I'm actually thinking of starting a website to create "comic-style" (as the author calls them) explanations of important papers and ideas in science. Would anyone here be interested in writing or explaining papers or ideas? My email's in my profile. Even just doing a bit of illustration or writing would help.

I'm also trying to figure out papers and ideas that are important. I'd like to also include some important things that aren't well explained on the internet already, or are only well known within a specific field. Anyone have any suggestions?

> I'm actually thinking of starting a website to create "comic-style" (as the author calls them) explanations of important papers and ideas in science.

There already is a hub for such projects: http://explorabl.es

Perhaps you can get some inspiration and/or exposure through it.

That's an interesting site but doesn't seem to be picking papers the first one I picked up on was the Monty Hall game show problem which isn't a paper as far as I know.

The Monty Hall problem was invented by Steve Selvin in a letter to The American Statistician journal, and it's only inspired in the game show.


No doubt this “sequential art” and “comic-like format” notion was at least partially inspired by Scott McCloud’s “Understanding Comics”[1]. By Bret’s own admission, the book is one of the works that had “an extraordinary effect on [his] life or way of thinking”[2].

I agree. That book is phenomenal.

[1]: http://scottmccloud.com/2-print/1-uc/

[2]: http://worrydream.com/#!/Links

I love it when people do take the time to do good information visualisation, especially when it helps everyone understand a complex idea.

However, I do think that those who believe that all academic papers should take this form fundamentally misunderstand the purpose of academic papers. Sadly, most of these papers are not about improving universal knowledge - instead they are simply attempting to convey to their peers some insight into a commonly understood problem or question. Because they are aiming at such a small target audience who have a deep understanding of the problem domain, shorthand and skimping on explainatory visualisations is not only an acceptable choice - it can make the difference between a paper being published or not. Few academics have the time, experience or funding to make beautiful visualisations - and those that can afford to spend them on that aspect often do so in exchange for depth of insight - and thus they are often seen as less valuable.

This is consistent with my experience. Many papers seem to rely on unstated conventions regarding the problem definition and use of variables. Say I'm not in the "small target audience who have a deep understanding of the problem domain," but would like to read and understand a paper. What's the usual approach to overcome this barrier?

You'll most likely need to get acquainted with a significant chunk of the literature in that domain. For the most part, papers are self-contained, but they build on top of past work.

The best way is to work your way through the references cited. Often you'll find a few that come up again and again. Read them, and you'll have gained a bit more understanding on the topic. Rinse and repeat until you do understand the problem and how it is commonly treated in the literature.

Another important distinction is that absolutely not everything is even close to visualizable in short sweet diagrams.

Wouldn't it be nice if research teams were able to publish their work in easily accessible html5 interactive applications? Such as ConvNetJS for no sweat browser based CNNs:


But even with libraries such as D3.js for visualization. Or Tensorflow.js for WebGL accelerated learning. Many scientists simply lack the front end skills necessary to create usable demos. Which can mean an opportunity to reach out to a team if you are looking for projects to work on.


We can also see where this is ultimately headed with the rise of electronic lab "notebooks" such as JupyterHub for data science. We want an intelligent interface, that could even understand human language to design experiments. It needs to be scalable and distributable to large enterprise research teams spanning multiple locations. Results have to be repeatable, obviously. And with a single click, publishable to the entire scientific community and general public.

What the future looks like is science "as a service"

> Wouldn't it be nice if research teams were able to publish their work in easily accessible html5 interactive applications?

Sounds nice. Who pays for this? Are you willing to trade less science for more html5?

> What the future looks like is science "as a service"

It already is and always has been. Almost all science results are openly available to read. The most significant bits are translated into digestible textbooks within a handful of years.

jesus fuck no.....

"it needs to be scalable and distributable to large enterprise research teams spanning multiple locations."

'What the future looks like is science "as a service"' the day this happens is the day I stop studying TCS and start working on autonomous weapons. If we are going to let the world burn I might as well have fun participating..

Sure and that's why we have science magazines like New Scientist, National Geographic, Nature etc. They read papers most of us won't understand and translate them into a form that is more approachable.

I would't want the actual scientist to also have to worry about the "design representation" of their research. If their work is important enough, others will do it for them.

> I would't want the actual scientist to also have to worry about the "design representation" of their research. If their work is important enough, others will do it for them.

In reality though, that barely ever happens.

More like they read papers most of us won't understand and deliberately misinterpret it to create catchy headlines.

New Scientist is awful

So people have been saying for the last 30 years. Yet no one seems to care enough to publish something better. Or even to suggest an alternative.

I think this sort of things are good for scientific divulgation, for the general public, but a poor choice for communication between scientists. The design choices that make for a great popular science article make for a bad scientific paper. Plus, many ideas simply aren't amenable to simple pictorial explanation like this one. Good effort nonetheless, but not something of general applicability.

There was a recent interesting paper on the arXiv which argued that the ubiquity of small-world networks was vastly overstated.

Can't find the paper anymore - did anyone else see this / have a link?

I remember there was a paper regarding scale-free networks. I believe it was discussed on HN a few months back:


But I also (perhaps falsely) remember a discussion that seemed to challenge the small-world networks as well, so that may not be the discussion I was thinking of. I may not have engaged my brain at the time but I'd be surprised if it was small-world networks given the widely-reproduced evidence for the effect.

Yes! That is the paper I had in mind! Thanks.

I will add that organic chemistry regualarly uses sequential art in the form of chemical reaction mechanisms.


This should have a (2011) tag.

Thanks! Updated.

I'm not sure what point Bret is trying to make here. Of course interactive explanation can be powerful. As a complement to the actual paper, this can be very valuable, similar to how the HTML version of a paper is useful in many contexts.

But every scientific paper ever published will be around until the end of humanity, whereas this interactive web page will surely disappear within a decade or two. The scientific paper - a self-contained entity that can be moved around, printed out, archived, searched and annotated - is an incredibly powerful entity.

> whereas this interactive web page will surely disappear within a decade or two

The point you're raising is at the moment important, but ultimately irrelevant.

The scientific paper in digital format when it was first produced, did not have all the advantages of physical paper. Physical paper was easier to archive (digital storage would corrupt easily), easier to move around (you needed a computer to view a digital paper, and most people didn't have one; networks were mostly non-existent), and easier to annotate.

These were all important things to consider at the time, but the technology eventually caught up, and surpassed printed paper in most of not all aspects.

Keeping dynamic content consumable through the years, would be costlier than keeping static content consumable. But the price isn't that high. Web standards are designed with backward compatibility in mind, and the software you use to view web content (browsers) is mostly open source. I'd imagine it'd be much easier to view web content produced today in the 2030s, than it is for us to play NES games produced in the 80s.

You are correct but is not enough for the technology to improve, there needs to be some thought into making things forwards compatible too. Human readable formats stand the best chance of being understood even if there is an interruption in the advancement of civilisation. Paper obviously wins here. The elephant in the room is of course the media on which the information is stored. Paper is probably the most robust storage technology we have right now.

Seems like many in this thread are thinking about conservation. Thats positive.

Not all paper (papyrus, wood or stone) based documents remains till this day. Effort was made to conserve it because, through different times, enough people thought that knownledge was worth being remembered.

If a interactive paper is meaningful to enough people, it will be preserved.

You could say the same of oral traditions, which were also preserved: but that doesn't mean that oral traditions are just as good (in terms of information transfer) as written ones. Keeping up a digital paper would be much like an oral tradition, with each new generation having to re-implement it for current devices.

I have books that were printed in the '70s and '80s that are yellowed and cracking--literally disintegrating where they sit, on a bookshelf in a climate controlled room. You certainly can produce and print on paper that lasts much longer, but this is also true of digital mediums.

There was one NES, some of it's contemporaries lack good emulators. There are many web standards, and not all of them are backward compatible.

> every scientific paper ever published will be around until the end of humanity, whereas this interactive web page will surely disappear within a decade or two.

I think you're conflating two distinct issues. One is whether the media is static or interactive, and the other is whether it is a page on the web or stored in some other fashion.

If the scientific papers will be around long-term, that's likely because they'll be stored in some sort of database. There's no reason why interactive media couldn't be similarly stored.

(Yes, there are issues about being able to run the interactive content in the future, though you don't bring that up).

The primary point of a paper is to disseminate information and communicate progress, not archiving learnings. So his point is that the primary purpose of a paper is better served using modern media.

Imagine what effects could cascade throughout society if the slope of the learning curve to master a paper was reduced just a tiny amount. Sure, regular everyday people might not be directly affected by it. But what it would mean is more researchers understanding more papers at a deeper level, and faster since the effort per unit of time invested will have been significantly reduced. It's true that websites have a expiration date. But this is one of those low-hanging fruits where the technology is perfectly mature and the problem is obvious. And the combination of the two could produce massive productivity gains, with relatively little effort, if you just made what Bret is suggesting an habitual norm within academia.

> It's true that websites have a expiration date

I’m not so sure about that. Web standards have made it so that the browser you install today can still view the websites that were made when your favorite Super Nintendo game came out.

And you wouldn’t print just one copy of a paper ever, so a network of database servers getting destroyed is the equivalent of a bunch of physical libraries getting destroyed.

Flash may have gone away but Flash was not a core component of the web spec, like HTML. It was a third-party plug-in.

It would be the equivalent of creating more random shorts in the graph describing access to new ideas. A more efficient small world network. I have to assume the paper used was chosen for just this reason. I really enjoy that.

Probably the most straightforward way of getting additional information beyond the format of the scientific paper is to have additional supplementary videos. I'm fortunate to be in a field (computer graphics) where there's a culture of having submission videos alongside the manuscript.

In a world where the PDFs are being backed up, the videos are also usually backed up and are just as available. I think in most cases, having an animation gets most of the value that you'd get from having complete interactivity.

An HTML representation is still self-contained in that it can gracefully degrade to mere text and images.

«60% of my fav links from 10 yrs ago are 404. I wonder if Library of Congress expects 60% of their collection to go up in smoke every decade». And the discussion that follows about how to preserve not just the bits but also their meanings: https://twitter.com/worrydream/status/478087637031325697

> whereas this interactive webpage will surely disappear within a decade or two.

If it disappears it will be because nobody kept it up. Not because your web browser is incapable of rendering them in a decade or two.

An interactive webpage can be self-contained. It doesn’t have to load cross-origin resources, it can simply include them in the directory. And if it’s a scientific paper it should have no need to load data from external sources. An interactive scientific paper can absolutely be self-contained.

Anyways back to the main topic, if it disappears in a decade or two it will be because nobody kept it up. For example if it was hosted on some random guy’s domain who eventually didn’t feel like hosting it anymore, etc. But even physical paper-format scientific papers aren’t just stored in random people’s garages, they are also archived by other organizations. So if archives like the library of congress kept their own copies, then you would expect the content to last as long as the library of congress (or other archives) would want them to.

A genetic mutation eventually becomes present in all of the population or in none. See: https://en.m.wikipedia.org/wiki/Fixation_(population_genetic...

For similar reasons, given enough time I'd expect results from a paper to be in text books and common knowledge, or be forgotten. Either way nobody would learn it from the paper after a few decades.

My experience is the exact opposite. I've regularly pulled data/observations/ideas from 1920's-1960's era articles and cited them. (Heck, I've used depth soundings from Capitan Cook's expeditions in the 1700's as data when making detailed bathymetric maps -- it was the only thing available in the area.)

You can't be a researcher in a field and only read articles from the last 20 years. I've lost count of the number of times I've been struggling with a problem, come up with a partial solution, then found a nice, complete, well-thought-out solution in some random engineering journal from the 40's. Use the literature. Never, ever, ignore papers just because they're from a few years back. You may need to filter the ideas through a few paradigm shifts, but the underlying observations and methods can be very useful.

Textbooks can be wonderful overviews and great ways to learn, but they only scratch the surface of topics. Furthermore, in most cases, you can't directly cite textbooks. Only original research is citeable. Review papers, most textbooks, and other compendiums are not citeable in any field I've been involved with. The exceptions are textbooks that have original research in them, which are very common, but are more akin to special editions of a journal than most textbooks.

Good post. I'm disappointed by how many researchers don't read papers older than 10 years or so. Too many people seem to think that more recent research supercedes previous research in every way, which is not true.

This often leads to researchers reinventing the wheel. Good ideas can fall out of favor essentially randomly for indeterminate amounts of time, so there should be no expectation that what is cited today is necessarily the best.

To use another example I've seen a few times, an approach that might have seemed like a dead end in the 50s could now be tractable due to other developments, like modern computers.

Citations are different, for instance I'm a research bioinformatician writing a paper about the issues you face storing billions of genotypes in an SQL database.

I'm going to cite Codd 1970 but I learned SQL from web tutorials written 30 years after the paper, which I've only skimmed.

And of course, Chrome on Android uselessly offers "Simplified View" which breaks the new visual flow of the document.

perhaps the point is also 'code to learn' - i always thought archiving these constructions would be an interesting way of seeing how tangentially educated people can get themselves from their area of expertise to this result and a fantastic resource for students or life long learners.


Speaking of Victor, this was just posted today:


HN article:


Was this grey text on a white background for most people? I found that it made me squint rather.

It's unreadable for me. I suspect all "grey text is easier o read" people either have bad monitors or wrongly configure them.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact