notes on interactive visualization, optimizing for user understanding and the ominous creative technologists

-----------------
i had a lot of skepticism for the field of human computer interaction? like what does it mean? a field dedicated to optimization (going from 1 --> 2, instead of 0 ---> 1 /s). hci researchers or more popularly creative technologists. vernacular almost as aesthetic as their webpages. building tools for thought that enable human agency?

i don't know, i read their words, on their digital gardens and think they must be the cool, indie kids of the tech industry. words housed on their digital gardens. it was a field that i never even interacted with before, but the twitter algorithm acted on god's will once again, and boom...new rabbithole. [*jacky*](https://jzhao.xyz/), [*spencer*](https://www.spencerchang.me/), [*andy*](https://andymatuschak.org/) and [*maggie* ](https://maggieappleton.com/) were my quintessentials. at first i interacted with this breed of work the way i do physics that i don't understand: fetishized, amazed to the extent that i assume all things difficult are superior and questionable in that i couldn't piece together how these people get paid, and pretty handsomely as far as i know.

The last point is more negligible now. 2/3 that i mention are employed at large tech companies and the other used to work at one.

right now i'm writing on obsidian and i like it a lot. i'm still the person that uses google docs, the notes app (not synced across my devices :skull ), etc, but obsidian is my main writing tool at the moment and its been great, i still use the others, but it's really enjoyable for anything longer than a hundred words. a non-linear approach to keeping store of my files. akin to the hypertext format established somewhat by ted nelson in literary machines. on a very very simplistic form of this, it's part of the reason why i like doing dives like this, chock-filled with backlinks and am a huge fan of [curius](https://curius.app/hamidah-oderinwale). look at what i'm thinking about, interested in, on multiple dimensions: the thoughts that sparked the rabbithole, the time they were had and what they led to.

turns out the op intuitive designer folk, deserve the hype. it goes beyond prettiness, it's about usefulness, ease. based on the premise that a world where we work (slightly) better, is a better world overall

i sound ignorant, but when i started pulling on this thread i was far from letting the last point stop me from thinking that at some point i felt as though i didn't understand, and more concretely i couldn't articulate the theory of change. it's okay to stop at "this is cool", but it feels really unsatisfying, and trying to understand everything is difficult and not worth it, insofar as our time is finite, but i guess the algorithm won't let me go, so i digress.

--

with the first thing, which is that i saw no "problems". i open up google, and the magic lies in the ranking algorithm, not the ui. right? i go to the site of any computer illiterate, but still [*cracked economist*](http://web.mit.edu/krugman/www/) (my academics of choice), and you best believe their site is just pure html, minimal css. times new roman. font 12, some not even putting in the effort to give us some padding. i don't really care though. i read their blogs, feel unjustly intellectual, and go on to the next.

as you can imagine, in the process of educating myself, i realize most of the above isn't true. there are interesting things in this space, and some re-connection with thoughtfulness as a virtue to appreciate, made some sector of the efforts of the creative technologists feel really meaningful. i don't speak for the entire field. i'm not part of it. i don't know enough, and i still think lightweight personal sites reign supreme, but let's delve in a little further, shall we.

----

the key value add for people who work on these things are making "tools, resources, etc." more accessible. chris olah, started [*distill*](/https://distill.pub/) - a machine learning journal that focuses on making research in the field "clear, dynamic, and vivid". it's very solid. to the extent that i need to understand their paper on [*understanding convolutions on graphs.*](https://github.com/distillpub/post--understanding-gnns/issues?q=is%3Aissue+label%3Apeer-review "View this article's reviews as Github issues.")

pretty. interactive visualizations seems much better at building the intuitions necessary to understand well. distill articulates their impact and in turn verbalizes the greater idea of thinking about the user, reader, etc. of course, "clear writing benefits everyone" as they lay out. jargon is a chrome extension that aimed to work on just that: [*converting legalese to plain english*.](https:/adaobiadibe23.medium.com/jargon-81c0c25530d9research) debt is a thing and you're a new researcher trying to build foundational knowledge and you have so many things to possibly read, the papers themselves are dense, it takes time to acclimate yourself with the needed venacular, etc. distill explains how the "climb isn't progress: the climb is a mountain of debt". this sort of makes sense.

to the extent to which we're trying to stretch the importance of connectivity for the greater benefits of human welfare. there's an interesting story of three researchers who emailed terrence tao upon being unsure of whether they'd discovered

ultimately, with your knowledge as a utile (bah): the more it increases the more net positive you are, but when you think about the energy you have to expend to get to the final product: spending time grovelling through the insane amount of info available, checking for your own comprehension along the way and assessing how it all fits into your big better is all energy spent not actually making progress on the thing you set out to do. i don't believe this with that much confidence. i don't think i'm grabbing at the thesis here, but for the most part i get the sentiment. distill is working on eliminating this debt.

distill feels like a product. a focus on practicality. very intentional in thinking about what the user can gain from the journal, even though ultimately if you're looking to read ML papers, the research would be considered the main thing to gain, regardless. so, we have this paper on "understanding convolutions on graphs", it's meant to be explanatory and in turn does a really good job. i won't delve too much into the writing and the explanations although i think their the most redeeming quality. the interactive graphs are a major highlight.

to help readers better understand "polynomial-based" convolutions, they have this interactive graphic. with sliders you can change the values for the terms (x) {vectors of the coefficients} which is the polynomial of the form. affect (x'), the result of the convolution.


very very dope. it's unrealistic to expect this for every technical paper, especially where gray areas in understanding exist and you're dealing with ideas with levels up in abstraction high enough, where coming up with intuitive graphics is just really hard, but ditching [PDFs for scrollable interfaces](https://worldmodels.github.io/), intentional headers and runnable code (where possible) is an idealistic, but chef's kiss society.

unfortuantely distill has been put on hiatus, makes sense, this seems super time-consuming. thinking about how you want your work to be recieved is tiring, is a super iterative process and is especially hard when you're dealing with very technical concepts that have yet to be full understood themselves. it was ran on a volunteer basis by people with time-consuming full-time jobs, optimizing for accessibility doesn't directly bring the money home. distill described as a "beautiful artifact", feels very true.

----

in trying to synthesize what i like about these two examples, i guess i'm most excited about the applications in education when looking at "user research". you want the endpoint to get as much as possible from your resource, and you optimize accordingly. the content still has to do the heavy-lifting and you can't write bullshit, but the extra mile goes a long way?

the cornerstone piece from what's being produced in this area of work, could probably be "how can we develop transformative tools for thought" by matuschack and nielsen, ideating a medium that makes it easier for people to remember what they read.

the basis for this is the mnemonic medium. remember mnemonics? using your knuckles to remember the number of days in the months of the year, or remembering that "good bikes don't fall apart" for music letter names in bass clef: m and n go deeper, thinking about how this applies to the essay medium. they don't incorporate mnemonics in the way i exemplify above, but rather in the more general sense of helping you remember. they lay out an essay of sorts explaining quantum computing with questions nestled throughout. questions are interactive and they get to check their own understanding as they pop off. if you selected "didn't remember" for a fact when it is tested, then the time interval for when it pops up adjusts as a result, similar to Anki as an example. It exploits on the central principle that the more repeatedly we are tested on knowing a concept, the stronger our memory of it becomes.

it goes beyond say, anki, in that it claims to help people improve their learning of the abstract, building conceptual knowledge, etc. out of scope for traditional memory systems. this seems especially powerful, in that pursuing novel work of most kinds in most disciplines, requires that you both understand the larger picture and the intricacies of the area.

----

i guess something about the field is that it doesn't have the romance? anyways i think jeff raskin helps rekindle this. we hear about steve jobs being a product guy and not an engineer and how this led to their success. this intentionality. the technical, mumble under their breadth, but its a narrative that satisfies some desire for a cute story. from his wikipedia page, he also seemed super wholesome.

jeff raskin was a so-called "user-interface pioneer", helping with the design of the first macintosh computer, and then attempting to set out on his own to work on *archy,* where he would make word processors and email inboxes more intuitive. he wrote the book, the human interface, which i haven't read, but is pretty famous? he lays out some sort of design constiution that at its core doesn't actually fit the whole "enable human agency" bill.

the book conceptualizes a world where friction is eliminated, things are effortless, habituation is built and the user's mind is freed. understanding this, helps put this new wave of hci cool kids into perspective. they try to research and design back human control. raskin's son, aza, co-founded the center for human technology, the hallmark of the film, the social dillemma, which sought out to shine light on human dependability on addictive tech systems. a dependability that has everything from human connection to our democracy, at least as it was argued in the film.

although there are explicit efforts to work against surveillance within a totalitarian government lense, there are more discrete forms of disablement that people are working against? surveillance capitalism, a loss of autonomy at the mercy of z algorithm, etc. as per a commandment suggested by christoph labacher:

> **Good digital design gives users the freedom to take their data and leave.**

bah this all kind of goes back to my 1-week phase of getting into berlin history and related tangents. the warner era was just intense. christopher went to the ulm school of design which being founded in the late 1960s, just seems(ed) super dope: a technology focus, an institutional heartbeat of reclaiming expressionism as a tool for defiance (this is me being mawkish rip), and intense interdisciplinarity.

------

matuschak used to work at khan academy labs which when thinking about learning maxxing probably takes the crown, it makes sense he focused on this topic and it makes sense its as robust as it is. nielsen is a polymath of sorts, a scientist at his core, he writes about openscience and digital pedagogical tools, as well? they go off and make quantum country which essentially helps put the theory laid out in TTFT into practice ( :DDD)

----
i guess things start getting a little more interesting and also confusing when you look at the theorists trying to delve into how we interact with computing, the internet, etc. the internet is mighty, much information to interact with: without the specific use cases of education, i found it much harder to wrap my mind around some of the work done in the space.

more generally when i think about clarity in expression.

its very easy to romanticize jargon. it makes you feel smart. being especially verbose [(cough, cough)](https://www.hamidah.me/blog/). only those who are "smart" can understand what you say, you feel better about yourself. those "people" aren't smart because they talk like that, maybe they're brighter than average because they spent the time trying to pick it up? paul graham is the quintessential reference when you think about writing clearly, at least according to twitter. he studied philosophy after all, where succintness is king, so makes sense.

i look to spotify, and maybe its just because my recent playlists are of the tearjerking flavor, but like damn they (as in the company with almost 10k employeed) took so much to think about me lmao

yes, i know, i'm just another set of digits forking up 9.99 monthly (because quebec doesn't allow it's residents to take advantage of their promotions :(( ), but the app is beautiful, i get solid recs from artists that quickly become favourites, i have choice and everything i could want is there. you thought about what would make a good experience before i even had to think.

woah this is super dope. i have literally only touched the tip of this weird iceberg, but everything: all the websites, the drivers of doing this in the first place and the actual output has been super fun to look at, to the extent i've explored a decent amount. i'm particularly excited about looking into knowledge sharing as it relates to academic papers: seeing how the premise of distill can manifest into more creative applications, finding more engineering types who fit this bill of super intentional and what their processes look like for actually getting things done, and hopefully looking into some of the key literature that these people have to offer : )

links. have fun!!!
- https://worldmodels.github.io/
- [distill ](https://https://distill.pub/)
- [arXiv](https://arxiv.org/) <333 (recently learned it was pronounced "archive". we love to see it :) )
- [elicit!!](https://https://elicit.org/)
- [quantum country ](https://quantum.country/)
- https://christophlabacher.com/notes
- https://jzhao.xyz/
- https://topos.institute/networked-mathematics (relevant to making research things more accessible)