Twitter Hashtags: The Death of an Affordance?
When I wrote my PhD between 2010-2014, I was, in part, interested in exploring how Facebook was constructed as a place. To do this, I borrowed from the theories of urban sociology and human geography. In my thesis, I wrote:
While the architecture of Facebook has certain objective limits, it is possible to go beyond, disrupt or reinterpret these determinants that Facebook – as an object – sets on its utilisation (de Certeau 1984). De Certeau argues that the act of walking through a space (the urban environment) creates and transforms spatial signifiers into something else. The choices made by the walker mean that only a few possibilities out of a constructed order are realised; conversely the number of possibilities may be increased by taking unexpected routes and limited through prohibitions. That is, taking or not taking paths considered accessible or obligatory (de Certeau 1984). Thought of in this way, Facebook could be considered to have as many forms as it has users (de Certeau 1984).
By making these choices, places are made and unmade. The history of the internet, and social networking, is replete with such examples of previously well-trodden paths becoming neglected, and then abandoned, because the spatial choices made (of where to go) have condemned them to inertia and disappearance.
I am now interested in how this ‘inertia and disappearance’, this not choosing, or not walking is playing out with hashtags on Twitter.
In the early iterations of Twitter, hashtags functioned as a way of curating information, and emerged somewhat organically as a way of working with and around the early constraints of the platform.
The first use of Twitter hashtags was the 2007 San Diego wildfires. As Bruns and Burgess explain, this event provided a clear ‘use case’ for hashtags on Twitter. To circle back to the metaphor of walking, a ‘goat path’ or ‘desire lines’ emerged through the architecture of Twitter, opening up new patterns of use.
Hashtags became a central part of the Twitter experience. Hashtags have been used for live TV watching, elections, social issues, and natural disasters.
This was also demonstrated by the use of Twitter during the 2010-2011 Queensland floods via the #qldfloods and #bnefloods hashtags which played a central role in disaster communication and recovery coordination. Hashtagged information on Twitter then flowed into broader media coverage of the floods – expanding the reach of Tweets beyond their initial audience and platform.
Gathering data from hashtags also became central to researchers’ work to understand and map Twitter dynamics. Why does this story of floods and hashtags matter?
It matters because I think the hashtag is dead, as a meaningful Twitter affordance. Here I draw from Jenny Davis’ mechanisms and conditions framework of affordances, which I think is very useful for the argument I am advancing here:
[using a] mechanisms and conditions framework shifts affordances’ orienting question from what technologies afford, to how technologies afford, for whom and under what circumstances? The ‘how’ of affordances, or its mechanisms, indicate that technologies request, demand, encourage, discourage, refuse, and allow social action, conditioned on individual and contextual variables, grouped into perception—what a subject perceives of an object, dexterity—one’s capacity to operate the object, and cultural and institutional legitimacy—the social support, or lack thereof, for technological engagement.
Importantly, Davis’ work highlights how affordances are co-constituted through both technical capacity and social patterns of use.
Don’t get me wrong, hashtags still exist, and commonly coalesce around significant social events, like elections, sporting events and transformative social issues like #BLM. However, they have fallen out of use and style as an everyday communicative practise on Twitter.
To illustrate, we return to Brisbane, which experienced widespread flooding in March 2022 due to intense and prolonged rainfall. While the floods were not as significant as those in 2011, there was still widespread major flooding along the Brisbane River and in other creek and river catchments throughout Brisbane and South East Queensland.
As the disaster unfolded, I turned to Twitter and immediately began searching the #qldfloods and #bnefloods hashtags. At best, they returned a little over 100 results each. Highly centralised broadcast media in Australia was not particularly responsive to the unfolding disaster. The ABC had a good live blog, and their local radio stations did their usual excellent emergency work, but commercial stations were largely silent as were the Twitter hashtags.
Concerned about the impending flooding, I texted a friend who I knew lived in a flood-prone area. She responded with surprise; all the news she’d seen was about the war in Ukraine. I told her what I knew and directed her to the ABC live blog.
On Twitter, Professor Jean Burgess also noted the failures in the local media ecology.
Any information I found about the Queensland floods on Twitter became a function of the algorithm and socially-proximate networks of retweeting. If you sat outside those retweeting networks, or the algorithm did not surface those tweets for you, information was thin on the ground and difficult for a casual user to find.
Significantly, reporters, who in the past would have used hashtags to circulate flood updates did not use them. Twitter users, filming their slowly flooding streets and backyards also rarely used hashtags to mark their content as flood-related. The combination of a diminished local media system, and the haphazard curation and sharing of flood content on Twitter, meant it was very hard to get a sense of what was actually happening in anything approaching real-time – once Twitter’s greatest strength as a social media platform.
The reasons for the functional death of hashtags as a means of curating information on Twitter are twofold and bring us back to Davis’ work on affordances. Twitter, as a platform has significantly reshaped its user interface over the past few years. Timelines are no longer straightforwardly linear. Tweets are presented based on an algorithmic mixture that shows you likes and retweets from people you follow, as well as suggesting tweets associated with seemingly random topics. The image below was suggested to me as a topic to follow last week.
Even trending topics have had a considerable overhaul. The trending column of hashtags is now hidden on the Twitter web interface, replaced with a series of trending topics (including promoted ones) called “What’s Happening”.
Notable here is the emphasis on live events (war, The Grammys), which are similar to the initial affordances of hashtags, but with a notable difference. Results within the trending topics are now algorithmically curated. Most recent tweets no longer appear at the top, but rather tweets with the most attention. The remaining showcased tweets rarely use hashtags, but rather are grouped based on their keyword use. Generally, this works well enough but topics can quickly become junked up, or filled with irrelevant content, as the algorithm cannot discern the context surrounding the key phrases.
It's hard to know what came first in these contexts, the Twitter redesign that centres topics over hashtags, or the declining use of hashtags spurring a more topic-focused redesign. Either way, the mutually shaping effects of both on platform architecture means that hashtags, once a commonplace communicative practice on Twitter are becoming much less widely used outside of what we might call ‘exceptional’ cases. For example, large scale live events, or social movements. The lack of volume on hashtags has other flow-on effects. Research suggests that it is increasingly easy to make hashtags ‘trend’ even with a relatively small number of Tweets. This is an artefact of their disuse. As day-to-day use of hashtags has decreased, they have become easier to ‘weaponise’. Dr Tim Graham details how trending hashtags are subject to coordinated activity and manipulation (also called brigading) to make issues trend. Hashtags then can no longer be trusted to accurately reflect user engagement and interest.
The (almost) death of the hashtag as an everyday communicative affordance on Twitter has implications for researchers. 100s of papers have been written that approach Twitter hashtags very straightforwardly, as proxies for interest in and discussion of any given topic on Twitter. Given their lack, of use and marginalisation within Twitter’s own platform design, it may be time for a change in strategy.
In a recent paper on Twitter memes, which I wrote with Simon Copland, we adapted our data collection strategy to reflect the move away from hashtag use. The memes we studied in this paper were based on key phrases, for example, “a girl who struggled”, not hashtags. In fact, if we’d run hashtag-based queries looking for data, they would have returned nothing of relevance at all. I think this approach reflects a necessary change to Twitter research moving forward. As both user behaviour and platform architecture twist and change, so too must our research strategies. While collecting, cleaning and computing Twitter data based on keywords as opposed to hashtags is a more laborious process, I think it is a necessary part of accurately representing the shifting sands of Twitter discourse.
Sometimes, it takes a crisis to drive home the fact that something no longer functions as assumed. The recent Brisbane floods have highlighted the significant drift away from hashtags as a communicative affordance on Twitter, including by reporters feeding information back to the public. Traditional broadcast media is also reliant on Twitter to generate much of its content, functioning as a type of digital ‘vox pops’. Twitter is a way of providing a sense of what’s happening ‘on the ground’, which is then fed back into the Twitter ecosystem through news stories that cite Twitter.
This much is clear, Twitter and the way information is curated on it is changed, and has changed over the past few years. Increasingly, the way information flows on Twitter relies heavily on social proximity. What you get depends very much on what the people you follow are liking, retweeting and commenting on. Tapping into information through hashtags is no longer the dominant or even common mode of engagement. These changes happened slowly, gradually and incrementally through the push/pull of platform and users, but it has happened. There are new paths of use through the Twitter landscape, and the old highway of the hashtag is becoming quiet and overgrown. Now our research strategies need to change too, we need to go where the people are.
With thanks to Jenny Davis, for careful and critical eyes.