Social media platforms: what's wrong (part 1)

Social media platforms: what’s wrong (part 1)

Table of contents:

Problems with the platforms themselves:

1. The metrics social media platforms rely on DO NOT align with the metrics that provide a good social user experience

Social media platforms heavily rely on metrics such as Daily Active Users (DAUs) and user engagement (likes / upvotes / comments / submissions / etc) in order to entice investors that the platform can make a ton of money via targeted advertising.

What this means, however, is that this also heavily incentivizes the platform to ignore things like cheap, low quality or controversial content, because it drives engagement, and heavily incentivizes the platform to not care about “alt” user accounts or really banning users in general, because those behaviors still raise the amount of Daily Active Users and overall user count.1

This comic describes pretty succinctly my feelings on DAUs, engagement, etc.

The end result of chasing these metrics eventually becomes Dead Internet Theory – bots talking to bots talking to bots, bots generating fake content, bots starting arguments – all to drive engagement, which eventually the platform hopes advertisers will pay money to put ads on.

2. The “one size fits all” of social media ranking algorithms lead to problems and are never quite a perfect “fit”

One of the big controversies of current social media platforms today is that they “hide” behind Section 230 rules, saying they are just “distributors” of content but not “publishers”. But more and more people are becoming skeptical of those claims, as a sufficiently complicated algorithm that ranks content based on criteria it determines to be “good” vs “bad” seems very much like they are picking and choosing which content they want to show, which can lead them to feel a bit like “publishers”. 2

Additionally, the algorithms that are used to determine a user’s social media feed are largely deterministic – it may take dynamic things into account (the communities the person is a part of, what they’ve upvoted/liked in the past, etc), and the platform itself might allow for a small amount of sorting & filtering (e.g. sort by “newest” vs “hottest” vs “just my friends”), but in general the same “process” of determining the order and ranking of content applies to everyone, as engineering-wise this most likely helps with architecting / debugging / caching / giving everyone the same rough “experience” as everyone else.3

The downside of having roughly one “algorithm” to sort and filter content becomes apparent quickly though – both content creators of the platform, and the content consumers themselves, try to adapt themselves to the algorithm. 

Take Google search for example — although it’s not a social media platform, it is backed by an algorithm that weighs content from “content creators” (websites) and ranks that content accordingly. And although Google has tried with the best of intentions to curate search results for good content on the web, SEO experts have studied the way Google ranks content so closely that most users can no longer rely on their search results to find what they need anymore.

Ever wonder why a recipe for walnut soup needed a long-winded story about grandma fighting in World War Two before it got to the ingredients? Well, websites (at least at one time) saw that Google prioritized long-form content, and decided: “What could be more long-form than a pointless, rambling story before the user can get to the content they need?”

I just want dinner, not your life story…

Want recommendations for a good camping tent? Go to dozens of websites that “rank” the best tents, only to find that most of them are just barely disguised advertisements and affiliate links, and the site owner may be ranking the tents not based on how good they are, but how much they get from the referral.4

To work around this, users will add things to their Google search query such as site:reddit.com to get “real” opinions, a practice which has become so prevalent that Google even has partnered deeply with Reddit on this matter, but once this becomes the standard way people find content, that too will no doubt be infiltrated by SEO experts as well.

On YouTube we can see this too – Mr Beast, for example, has spent an enormous amount of time optimizing even his thumbnails to best get users to watch his videos, and all YouTube content creators know that to have success on YouTube they need to say things such as “Like, comment, subscribe” in all their videos. If you decide to go against the algorithm, many times you’ll be stuck growing your subscriber base very, very slowly, if at all:

A creator wrangling with trying to get the social media algorithm to notice him

From the content consumer side, there’s a good amount of effort spent trying to wrangle the algorithm to get it to “understand” them better – they’ll try downvoting things they don’t want to see, they’ll try to engage more with content they do want to see, but there’s a constant nagging feeling that while the algorithm will take some things the user wants to see into account, it will always have its own goals, whether the user likes it or not.

The Algorithm either knowing too much, or not enough

The concept of an algorithm defining how we engage with it (either as a creator or user) is not a new one – Goodhart’s law states:

“When a measure becomes a target, it ceases to be a good measure”.

So too does this hold true for social media platforms (and to an extent, search engines) – no matter what your good intentions are when creating your ranking algorithm (“We want to focus on long form content!” / “We want to increase smaller community engagement”), once people start to understand how the content is ranked, they will try to optimize for that ranking.

Up next:

“Part 2: Problems with the community aspect of social media platforms.“

Footnotes:

  1. When I say that platforms “don’t care” about these things, I should perhaps rephrase a bit — they do care, but usually only if it starts affecting their metrics or revenue. If they get a big backlash from their content creator community, for example, they’ll do just enough to satiate the community’s complaints, and no more. On the flipside, if advertisers start complaining, you’ll start seeing a platform do whatever it takes to fix the issue, no matter how small the complaint from the advertiser is. ↩︎
  2. I’m not trying to claim social media platforms are publishers, but instead trying to make the steel-man argument for the side that thinks they are publishers. ↩︎
  3. Note that I understand that the statement I present here is a bit too simple to be fully true — not only can algorithms be incredibly complex, I doubt they are necessarily always applied the same across users, and I similarly doubt that “debugging” is the core reason that they don’t have different algorithms per user. Still, it feels odd that more platforms don’t allow users to have fine-grain control over the algorithm that serves them content. ↩︎
  4. Seriously, there seems to be no ethics at all when it comes to product review websites. Even “professional” sites like NYTimes / NY Mag / CNN and more fill all their product recommendations with Amazon affiliate tags up the wazoo. If one of these products somehow could not be affiliate linked, would it ever make it into one of these lists? ↩︎

Leave a Reply