Over the weekend, social media expert Matt Navarra tweeted out this interesting tidbit based on info shared at the Online News Association’s 2018 conference.

Up and downvotes for tweets?

When you think about it, it actually makes a lot of sense - the up and downvote system used by Reddit, for example, is far better at surfacing relevant content and trends than the algorithms of Facebook and Twitter have ever got close to, at least at this stage. 

The difference lies in the human element - while Facebook, for example, has tried several times to use its algorithms to highlight more relevant content to users, and improve discovery, the matches they've shown tend to be unfocused, or crowded with spam and/or repeated posts. 

For example, back in 2016, Facebook ran a test of topic-specific alternate news feeds to highlight more content from across its network.

But that didn't really work out, because the topic matches were, as noted, very broad, unfocused, and irrelevant to most users - and if the content they highlight isn't compelling, or worse, filled with spam and junk, users are less likely to check back and see what else is on offer next time they're browsing within the app.

More recently, Zuck and Co. have come up with another, similar tool, which uncovers likely relevant content based on your interests if you get to the end of your News Feed.

But again, the matches are somewhat flawed, and it's difficult to tell what exactly they're based on.

Because Facebook's systems are reliant on algorithms, not human engagement, there are inherent difficulties in such discovery process. The use of algorithms also means users are less likely to be exposed to content outside their established areas of interest - there's plenty of great posts you're likely to miss in this way because the system is based on your personal activity, not what's more broadly popular.

That's where Reddit wins out. In the Reddit app, if you click on the 'Popular' tab, you get a great overview of the key trends of the day, exposing you to a wide range of interesting content.

Most of the posts there are likely not related to topics you've registered a previous interest in, but because they can be both up and downvoted, the system lends itself to uncovering the most relevant updates, which are also subject to human moderation within each subreddit.

Of course, Facebook's not likely to give users a 'downvote' option, something that Zuckerberg himself has long held reservations about. The concern with downvotes on Facebook is that they could be used as a negative element in terms of cyber-bullying, which is a valid and significant concern. While Reddit's system works well for that platform, Reddit also doesn't see anywhere near the usage that Facebook does, and Reddit's more focused on content, as opposed to the personal updates and interactions on Facebook.

Given these concerns, it makes sense for Facebook to have some hesitancy around a possible downvote tool - but they have recently started testing something similar within comments, though it's more focused on weeding out spam and offensive commentary than it is a vote on comment quality.  

 And now we have word that Twitter has at least considered something similar. 

Is that a concession that the algorithm approach is less effective - that, at some point, human feedback is more powerful for discovery?

I'd say its fairly difficult to argue against that, though scaling such an approach presents its own challenges. But an up/down vote process for tweets could make a lot of sense.

Unlike Facebook, people probably tweet fewer personal updates and insights, which could make downvotes potentially less damaging, and thus, risky in that respect. And the process could also greatly help improve discovery - rather than trying to highlight relevant accounts to follow, or algorithmically selected topics you might like in Explore, Twitter could provide a list of the most upvoted (balanced against downvotes) tweets of the day, which would likely highlight more relevant, broader ranging trends.

As noted by Matt Navarra in his tweet (above), this may never get a public test, we may never hear anything more about it - it's a mock-up that they've checked out within Twitter.

But it makes a lot of sense, and could help improve discovery on the platform.