Can Content Get Smarter?
Nodes & Networks, Stories & People
Before we go ahead, let’s praise Medium for what it’s done extremely well with publishing, it’s removed advertisements from the mix. You may occasionally forget this since advertising isn’t seen on the platform. It’s one of the few spaces on the internet where you can browse ad free, and in today’s startup economy, that’s suicidal. Yet the platform continues to experiment and innovate around that challenge, or as Business Insider likes to quote, “throwing shit at the wall to see what sticks.”
I don’t usually pay attention to startup tabloids but Ev Williams recent announcement seems to have alarmed the circus;
“Medium is not a publishing tool. It’s a network.
A network of ideas that build off each other. And people. And GIFs.”
As a frequent Medium user but also a product designer working with machine learning, I’ve been thinking about this challenge. I’m also following up on a previous nightmare I wrote about, The Internet Thinks You’re An Automaton. I work on a design challenge similar to content surfacing everyday, although it’s a building’s energy data.
Machine learning automates complex patterns & tasks but
that doesn’t mean it’s outcomes work for people.
I’m inundated with content everyday, and before we get into platforms, networks and features, assume that the selfish lazy reader that I am, that I want only one thing.
1. I want to read exactly what I want to read
and I don’t want to see trash.
I think you’d agree with that, so let’s also make a second assumption about my cynical self.
2. Most self-publishing is trash.
Sorry if that offends the genteel but have you talked to people recently and are you being honest with yourself. Ask anyone about Tumblr, Facebook or Twitter and they’ll tell you it’s mostly trash, and not just advertising. There’s so much trash that we think the stink is normal. But that’s fine, we persevere as readers and writers because there is also great self-publishing that emerges from time to time.
That stink isn’t normal. I subscribe to a handful of print periodicals, Harper’s, Atlantic and Nautilus for their algorithms, nay editors and they don’t stink — sometimes they smell funny at first but that’s besides the point. While with the internet, there are places I visit and places I avoid. From talking to friends, Medium is mostly a place people want to visit, and mostly for Matter, The Nib, Re:form and The Message along with a handful of other independent publications, Human Parts, Culture Club and Absurdist of course — that last one is me. Sometimes I’ll find some unaffiliated gems too.
So the budget cuts and restructuring of Medium writers is justifiably alarming for most people. The loss of quality supported content will be noticeably felt but it also doesn’t mean that “Medium is turning into Tumblr,” or at least it doesn’t have to I hope. Speaking for myself and others;
We left the noise of Tumblr and Facebook
to broadcast a signal to likeminded readers on Medium.
Assuming we’re all in this game together and that this really isn’t a race to being the loudest, how can content platforms work well for us, that is I assume the point of technology. Each and every platform offer platitudes on how their algorithms work best to surface content matched to you — yet mostly what I hear from people are complaints. Astra Taylor puts it well;
“We are at risk of starving in the midst of plenty,
free culture, like cheap food, incurs hidden costs.”
That risk of starvation is trying to find a decent thing to read that isn’t clickbait, heavy handed or self promotional. The risk of starvation is trying to find something more filling than a twee listicle, one liner or cat GIF. I’ll give you a personal experience of starvation. Before Facebook changed the page post algorithms, I was regularly getting views in the hundreds, sometimes thousands. When that changed a year ago to favor promoted posts, views droped to double digits, even though my followers have grown and I’m still bitter about that one.
Now to give a sense of data fairness to this recent drought some Medium users have observed, myself included, I thought I’d quickly tally the posts on my Medium feed, which went like this, Responses (10), Trash (7), Irrelevant (10), Fair (13) and Excellent (8).
This is a bit useless because I hadn’t surveyed prior but qualitatively, in my head I’d lump trash and irrelevant together, while most “responses” don’t interest me all that much. So for simplicities sake, let’s say there’s about a 50:50 chance I’ll click something and then an even lower chance I’ll read something all the way through it. All this is just to say, I really wish that Re:form still existed, or that I saw Message in my feed more, or that the things I care about appeared more without me having to go in search of it.
Borrowing an analogy from Astra Taylor, it’s like the suburbs, you have to travel through miles of cookie cutter before you find a decent place for coffee. Are we designing suburbs with our algorithms? Have I also mentioned recently that I’ve come to hate infinite scroll, the mere interaction reminds me of the banality of my precious life in the internet epoch.
Having said all that, what I really want to talk about are solutions, for networks and nodes, designed for human experiences — or at least my experience.
≈
Table of Contents & Feeds
A table of contents presents thematically aligned content for you, the web equivalent being web issues, limiting as it were, it rewards the reader with a rolling focus. The feed however offers serendipity and the endless, opportunities for chance encounters against the chaos of randomness. They have their strengths, but one approach offers higher signal to noise and for that, I prefer a curated layout. Did I mention that I deleted my twitter app recently. There is an inherent hierarchy in curated layouts that simply doesn’t exist in feeds. In feeds everything is equal, except we know it isn’t.
Kudos goes to Content Curation.


≈
Tags, Networks & Trends
Related to curation is the idea of identifiers, in the form of tags versus the fluidity of network sciences, and the realms in between. Most platforms do it all really, Twitter offers Stream, Discover and Feed, Medium works similarly, and I don’t even know what goes on at Facebook. But what really works for me?
If a new tag is created in the ocean, doesn’t anyone catch it?

Tags are quite basic, not much has changed since #hashtags really but trending tags are something. Social networks are a bit better unless they’re optimized for monetization or mind control, in which case they are effectively evil. Trends should be the obvious winner except in an age of ubiquitous trash, they’re often false positives. Yes, those cat videos are playing to your human vulnerabilities but not mine, except for anything that says Design in the title. I still bet on my networks first, tags for specifics and trends for serendipity. I do like that Medium has ceded some control with the Tags You Follow feature.
Point being, I don’t think any single method works and ideally, ambient computing would sort out content from your real networks, tags and trends to something greater than a 50:50 click. And ideally at the end of an article, you’d find the perfectly positioned piece, either from my network or interests relating to my current read. Instead most footers just present us with randoms or worse, placements.
Am I crazy in thinking content can really be tailored to the user, forecasting for some cataloged preference and social group with room for some serendipity. I can tell you that #LifeLessons will never grab my interest, no matter if it’s trending. This reminds me of when I repeatedly see an ad for something I already just bought or how Nest thermostat really miffed their experience by algorithmically deciding temperatures for residents.
Kudos still goes to Networks, but machine learning has a ways to go.
≈
Comments & Chat
I’ve never understood comments, it seems like a guestbook that no one reads. Chats at least are easy to follow. Hence the reason I use twitter over facebook, or conversation over say bathroom graffiti. Exchange is really important to me. Comments are somewhat engaging but not really, to be honest, as a person of color I stay away from comment sections.

Also, I don’t care about the 107 others, and who is Mark Zuckerberg? Seriously though, why are people using responses like comments on Medium. Save it for the comments or write a real response. I’m a believer that comments are meant to be counted and not read. Want to start a real dialog, then write a response.
Kudos goes to Chat, or in Medium’s case, Responses.
Which really brings me to why I first started writing this piece three months ago. I’m sorry if it doesn’t really fit into a neat listicle or some twee venn diagram but I’m hoping that publishers and platishers that want to surface better content will:
- Try machine learning that caters to a reader’s preferred tastes and networks.
- Afford some control over how that algorithm applies to them rather than uniformly or force fed.
- Offer the slightest bit of content serendipity, forecasting for content well outside their sphere.
- Always respect the user and keep advertising to a minimum.
Perhaps then we’d get passed 50:50 click. I know it’s possible since I read nearly every article of my print magazine but I’ve scrolled past most content today — as I expect you’ll scroll past this.
Final Note: While I was a strong math student, statistics was my weakest subject but I’m still fairly certain that the way startups count views, measure tastes and present statistics on their machine learning is skewed false.