The Trouble With Turning Communities into Code

When we track and automate audience behaviour on digital networks, are we making their lives easier, or creating more problems?

Buzzfeed published a great interview this week with Chris Weatherall, the Twitter developer who implemented the retweet function. But he didn’t invent the retweet - Twitter had a lot of its core functionality invented by users, not the platform team itself. The first retweet was by Eric Rice in April 2007, who added his newly coined phrase ‘ReTweet’ before copy-and-pasting a tweet criticizing a Spin magazine article on social media:

As with the hashtag and @-replies, Twitter noticed this emerging behaviour, and coded it into their product.

Let’s stop for a minute and focus on this process, as it’s an example of one of the most critical new developments in media networks in the last few decades:

Twitter builds a service that lets audience freely communicate with each other. It has limited functionality, so users start inventing shortcuts and hacks to add features to the service. Twitter starts to notice and track this behaviour, then decides to include it as a product feature.

This is a good thing, right? There are many design books on how to focus on emerging user behaviours to get insights into their needs, and how this can make your product better. So, changing the process from manually copy and pasting a tweet into a new tweet and adding ‘ReTweet’ or ‘RT’ into a simple tap of a button is a good example of user-centred design practise. Isn’t it?

Perhaps not. In the Buzzfeed interview, Weatherall describes his regret at how the retweet button affected Twitter as a public space. He recalls how the Gamergate controversy of 2014 - in which female games journalists were subjected to horrifically offensive and sustained criticism and trolling - was the moment he started realising the negative effects of the retweet:

“It was very easy for [the Gamergaters] to brigade reputational harm on someone they didn't like. Ask any of the people who were targets at that time, retweeting helped them get a false picture of a person out there faster than they could respond. We didn't build a defense for that. We only built an offensive conduit […] It dawned on me that this was not some small subset of people acting aberrantly. This might be how people behave. And that scared me to death.”

I’m interested in Weatherall’s insight that the retweet button was an ‘offensive conduit’, for which Twitter had no defense. When the retweet button (and soon afterwards, Facebook’s ‘share’ button) were added, most commentary was hugely positive. These platforms were recognising the social behaviours of their audience and turning them into code, massively increasing their ability to amplify messages that are gaining popularity, regardless of the audience reach of the original creator.

No-one imagined at the time how this could have negative consequences. As Weatherall recalls, “Only two or three times did someone ask a broader and more interesting social question, which was, ‘What is getting shared?’ That almost never came up.” Twitter focused on users’ behaviours, but were blind, or at least uncurious, to the context and consequences of the behaviour they were automating.


Before the retweet and share buttons, it was hard for someone with a small network to get their content out into very large groups. Ideas and memes went viral, but the process involved a lot more work - they needed to be shared manually within communities, often getting remixed and developed as part of the work needed to break them out into larger networks.

Kenyatta Cheese, founder of NYC social agency EA1, has given a number of excellent talks about how communities remix and adapt memes as they spread. In fact, he was one of the founders of Know Your Meme, a site dedicated to tracking the rise, adoption and adaptation of memes across the internet.

In his brilliant talk ‘The Story of the Gif’, he describes how the technically limited Gif format ended up outlasting the more sophisticated Flash format, precisely because it made the work needed to create moving image memes more accessible to a wider audience. Gifs ability to created short loops of video into a highly compressed format, together with its compatibility with basic HTML, made it hugely popular as a format for turning short clips of popular culture into a form of communication. Ground Zero for this behaviour was Tumblr, which had two critical features that encouraged Gif usage:

“The first is that you can upload a group of photos into a single “photoset” which means that you can get really creative on the timing and juxtaposition of animated gifs.

The second is that the heart of Tumblr is the “reblog” which is more than just a retweet or a share — it allows you to add your own text and images in response and when you reblog it, your response is just as important as the original post. This facilitates discussion. And because reblogs show up on the original post as “notes” it means that you can actually see what other people have written in their responses.

So if someone were to watch the Disney movie Snow White, she may love it so much that she wants to share it with their followers on Tumblr so she creates a gifset. She gives it tags so that it’s findable by other people exploring topical content on tumblr. The post gets reblogged, perhaps with commentary on how much they also loved the movie and what it meant to them.”

Cheese’s insight into how Tumblr’s ‘reblog’ button differed from Twitter’s retweet and Facebook’s share buttons is really important. Tumblr encouraged user comments on reblogged posts, and tracked them on the original post. They made the conversation as important as the original content.

When Twitter hard-coded retweets into the platform, they made it harder for users to add comments to a retweet - it was easier to just share the tweet without context. This led to people starting to ‘quote tweet’, adding their own commentary before they retweeted, and eventually Twitter added this as an option to the retweet button.

But this encouraged, as Weatherall calls it, the ‘dunk mechanism’, in which a quote is used to slam or criticise the original tweet as well as amplify it. For some reason, by making the work of sharing within a community seamless and less visible, Twitter created the tools for global shaming and trolling. As Weatherall recalls thinking when he saw Twitter mobs use retweets to attack their targets - “We might just have handed a 4-year-old a loaded weapon.”


Before the rise of social networks, the ‘work’ done by audiences as they discuss popular culture used to be mostly invisible. In a documentary about the making of the cult British comedy show Blackadder, Richard Curtis talked about what it was like to make TV in the 1980s:

‘This was before the days of ratings, I still don’t know how many people watched any episode of Blackadder [the first series]. I remember I used to wander round Shepherd’s Bush, looking in people’s windows, particularly people with basement flats, to see whether or not anyone was watching. I was watching to see if anyone was watching Blackadder One, because one didn’t know whether or not it had been a success or otherwise.’

When I started work as a Commissioning Editor at the UK broadcaster Channel 4 in 2007, we used to have regular Thursday meetings to discuss the performance of our shows and our competitors over the past week. We’d get presentations from the audience research team about ratings, and discuss reviews in the press, but we didn’t have any direct feedback from the actual viewers themselves.

I vividly remember the first time, in 2009, that one of the other Commissioning Editors brought out her laptop to show tweets in response to the broadcast of her show the previous night. She was incredibly excited, as she could see in real time how the audience were discussing her show.

Within a year, Channel 4 had commissioned Seven Days, a show which added this real-time audience interaction into the reality show format. The conversations on Twitter and Facebook were not only visible to the reality show participants, they were used to influence who stayed and who left the show.

I should admit at this point that I was running the multi-platform team at Channel 4 during Seven Days, and the show was a disaster. We built a website that used real-time calls on Facebook data to visualise this conversation, and a coding error combined with a huge rush of audience attention meant that Facebook blocked our calls within the first ten minutes of the launch episode, crashing the show site and creating huge amounts of negative comments on Twitter and Facebook. The show never quite recovered.

But a few years later, another format, Gogglebox, made by the same production company as Seven Days, found a more productive way to show how we discuss and share popular culture. Gogglebox has a rolling cast of ordinary people, filmed on their sofas and armchairs, as they watch TV together. It makes vernacular art out of the jokes, gasps and asides that we all make when we watch TV together with our friends and family.

In the same way that Tumblr created a kind of equality between original content and the work done by communities to share it, Gogglebox presents our everyday discussions about Television on the same level as the content itself. Unlike Seven Days, it doesn’t try to automate these behaviours. Instead it amplifies not just the original work, but the work done by the audience to contextualise, criticise and share it, bringing them both to the same level of visibility.


I think social software designers can learn a lot from Tumblr and Gogglebox. Just because we can track and analyse audience behaviours around content on digital platforms, it doesn’t mean that we should automate them. If we do that, we run the risk of creating tools that amplify the behaviour, but in so doing, erase it at the same time.

The retweet and share buttons on Twitter and Facebook are, in essence, just volume controls. They give us the opportunity to push something to a larger audience with one button, but they don’t require us to explain why we’re doing this. If you only give people the tools to make something louder, eventually all you will hear is the noisy screech of feedback loops.

At the end of the Buzzfeed interview, Weatherall discusses ways in which Twitter could change the Retweet button, either by limiting the audience you can share to (Whatsapp have recently limited message forwarding in an attempt to reduce the impact of misinformation spreading) or by requiring users to actually click on and read a link in a tweet before they can retweet it.

But this feels like the wrong approach - it tries to limit users’ behaviours instead of understanding the importance of their context. I’m not sure if it’s possible to de-toxify the public spaces that Twitter and Facebook have created. By building tools that amplified sharing behaviours, but made invisible the ‘work’ that we do to contextualise our sharing, they prioritised speed over community. They chose to automate the user behaviours they were seeing, without understanding the important role that little moments of friction can play in building communities.

Automating social behaviours is a great strategy if you want to build the biggest and most data-rich advertising network that the world has ever seen. But if you give people the choice of spending their time hanging out in the middle of Times Square, or hanging out with their friends on their sofa, I’ll bet that in the long run, the sofa will always win.

Loading more posts…