7 Oct 2018

Public Metrics on Social Networks

I’m convinced that many of the problems with social networks are attributable primarily to the way in which they handle metrics. Data is routinely made public without consideration as to its utility, and as a direct result users are coached to use the services in ways counter to their own interests.

Taking Twitter as an example, consider the various things for which the user is presented a numerical value on their own profile: Tweets; Tweets & Replies; Followers; Following; Lists; Likes; Moments; Notifications; Direct Messages. For the most part these are purely informational with respect to one’s own account; only Notifications and Direct Messages may include calls to action. Now consider the categories of information available when viewing another user’s public account: Tweets; Following; Followers; Likes; Lists. Do any of these numbers, as they relate to another person’s account, have any value to you? Let’s take each one in turn:

  • Tweets: Reading an account’s public posts is obviously central to how Twitter functions, but there is no need to display a count of how many posts they’ve made overall.

  • Following: It may be of interest to you to see which accounts a person follows. I would argue that this should be something that each individual user of Twitter should have the ability to share or keep private. In either case, there is no good reason to display the number of accounts they follow.

  • Followers: The same is true here. You might be interested to see who follows an account, but this should be something the account holder chooses whether or not to share. Regardless, there is no need to display the number of followers an account has.

  • Likes: I can see no reason to display the number of posts an account holder has liked. As above with Following & Followers, this should be something the account holder can choose to reveal or not. I feel Twitter’s introduction of Bookmarks in February 2018 — whilst they certainly have valid alternative uses — was partly a tacit admission that Likes being public isn’t ideal.

  • Lists: Compared with the others the number next to Lists is of little consequence, but the reasons why are revealing. Firstly, Lists are an aspect of Twitter over which the user does have complete control with respect to their public / private status. Private Lists do not show on a user’s Profile, and are not included in the count of Lists. Secondly, there is no judgment associated with the number of Lists one maintains.

This last point brings us to the second part of the argument. Not only does the public nature of these numbers have no positive value, it is often detrimental to the functioning of the service. Social networks are a messy hybrid of communication, publication, and advertising. Nothing good comes of the lines between these functions being blurred or erased, and nothing contributes more to this problem than public metrics. When the count of one’s Tweets, Followers, and Likes is unalterably public, one feels some pressure to keep an eye on the numbers, and quite possibly alter one’s behaviour on the platform (to one degree or another) to change them in the manner one desires to see them change. There is an important distinction here: for example between ‘wanting more followers’ and ‘wanting to be seen to have more followers’. The former may be rooted in a number of motivations, some of them beneficent. The latter is only really attributable to anxiety over reputation, and does no one any good.

The effect of this is even more detrimental in the case of individual posts. Publicly displaying the number of replies, likes, and retweets that a tweet has received is the single worst element of Twitter. Based on a growing corpus of research (such as that by the Center for Humane Technology) it is increasingly clear that algorithmically-curated services amplify content with a strong emotional valence. Put simply, a post that elicits strong responses gets interacted with more than a neutral post. More people reply, share, like etc., and as a result the system learns that these posts are more valuable. The next step is that the system makes these posts more visible or prioritises them in other ways, which leads to yet more engagement. This is true of content with a positive emotional valence, but it is particularly the case with negative emotional valences. Nothing travels faster and further, or garners more attention on social networks than outrage.

And as much as we’d like to pin this entirely on the algorithms, we’re not blameless in the equation. The lesson we’ve learned from using these networks is the same one the algorithms have learned: in order to get more engagement with our posts, we should post more extreme content. If you want a lot of visibility for your posts it’s better to make them confrontational. If you want your opinion to be noticed it’s more effective to inflate it cartoonishly. And the way you receive feedback on whether it’s working or not is by keeping an eye on the numbers. How many replies; how many retweets; how many likes? What should I do differently next time to make those numbers go up?

Think of this in terms of Instagram, where Likes and Comments are also tabulated publicly. Here the extremes are less likely to be around emotionality, and more commonly to do with beauty or æsthetics. If you post a photo of yourself or your living room decor on a platform where the metrics are publicly viewable, you’re going to naturally be concerned with how it will be received. Not ‘do I like how I look’ but ‘will others like how I look’ in this picture? When the numbers come back and they’re not what you’d hoped for, you naturally make adjustments to how you will present yourself next time.

This is behaviour modification by way of public score-keeping. It’s one thing that you can see that your selfie only got 6 likes, but it’s infinitely worse that everyone else has that information too. There is no reason that the number of engagements with a post should be publicly viewable. The only thing these metrics track accurately is popularity, and that is the least reliable and most widely misinterpreted guide as to value. That something is popular does not make it good. That something is unpopular does not make it bad. That outrage is more virulent than reason does not make it a better response.

Imagine the effects of removing these numbers from public display. You can still find out how many people, and whom, liked your post if you so wish, but no one else can. The benefits of altering one’s behaviour or appearance for a future post are entirely removed, as no one will know whether one post is more popular than another. Think of the times that a news report relies on the volume of engagement with a social network post as an indicator of veracity. It’s inaccurate and potentially dangerous to do so, but it’s common practice — such is the ease with which we mistake ‘popular’ for ‘good’ or ‘true’. It’s a problem easily remedied if the metrics are made private.

These numbers might mean something to you as a user. It bears asking yourself what they mean to you and why. How much of it is because other people see them too? Do you look to the numbers displayed on other people’s accounts and posts as indicators of anything other than popularity? And what value do you place on raw popularity where that’s all that’s being measured?

The unfortunate element of this that should be mentioned is that anything and everything that increases traffic and engagement is by default a positive thing as far as the operators of social networks are concerned. These are free-to-use, advertising-supported platforms, and keeping as many people as possible engaged for as long as possible is their modus operandi. If pushing emotionally charged content is what accomplishes that goal, that’s what they’ll do. And if keeping all of these metrics publicly visible means that users spend more time on the service trying to get their numbers up, that’s a good thing for the advertisers and therefore for the service. As far as I can see, that is the primary obstacle to making this data private (either optionally or by default). I certainly can’t see the benefit to the user of it remaining public.