Twitter as a system of ambient journalism

The idea that gave this page its long afterlife is a simple one: a stream of short, public messages can work less like a news bulletin and more like an awareness system. The paper behind it — Twittering the News: The emergence of ambient journalism, presented by Alfred Hermida at the Future of Journalism conference in Cardiff in September 2009 — borrowed the term from computer science, where “ambient” describes broad, asynchronous, lightweight and always-on systems that let people hold a mental model of what is going on around them without reading every message.

The argument mattered because the numbers were moving fast. In the year before the talk, Twitter went from roughly 1.6 million accounts to 32 million, and the platform had already carried the load during the 2008 California wildfires, the 2008 US presidential election, the Mumbai attacks and the Iranian election protests of 2009. Newsrooms noticed: UK national newspapers ran 131 official accounts with just under 1.5 million followers between them, Sky News had appointed a correspondent whose brief was to scour the platform for stories, and services such as MuckRack aggregated journalists’ tweets.

The scepticism was just as loud. Maureen Dowd dismissed the service in the New York Times as “a toy for bored celebrities and high-school girls”, working journalists compared searching it for facts to “searching for medical advice in an online world of quacks and cures”, and one flat verdict quoted in the paper ran “Twitter? I won’t touch it. It’s all garbage.” The paper’s answer was not that the stream is reliable — it is that reliability is the wrong test for it. An awareness system tells you where to look; verification remains a separate discipline, done the way it has always been done.

The borrowed literature was specific rather than decorative. Awareness systems are defined as computer-mediated communication systems “intended to help people construct and maintain awareness of each other’s activities, context or status, even when participants are not co-located”, a line of research that started in computer-supported cooperative work. For awareness itself the paper took Chalmers’s definition, “the ongoing interpretation of representations”, which allows a short public message to count both as a representation of human activity and as an artefact. On that reading the value of the service does not sit in any single fragment, which may be insignificant or of limited validity on its own, but in the mental portrait a run of messages builds up over time.

Scale was the practical problem. During the Iranian protests of June 2009 the volume of messages mentioning Iran peaked at 221,774 in a single hour, against a normal flow of between 10,000 and 50,000, according to an analysis by an interdisciplinary group of researchers based in Cambridge, Massachusetts. The need to reduce, select and filter grows with that volume, and the paper’s inference was that part of the work for journalism lay in building the systems that do it. Two early attempts were named: Twitscoop, which graphed activity around words and topics in real time, and NowPublic’s Scan, piloted with The Vancouver Sun to search conversations by keyword and surface photographs, video and links to breaking news.

The other half of the argument was about conversation. Link-sharing turned the stream into something close to a newspaper compiled from what a person’s network happened to be reading that morning, and carried a picture of the network alongside the news. Research by Boyd and colleagues on retweeting supported the point: passing a message on creates a distributed conversation others can follow without joining it, so the “stream of messages provided by Twitter allows individuals to be peripherally aware of discussions without being contributors” — which is where the paper met Dan Gillmor’s idea of journalism as a conversation.

The closing proposal was a role rather than a forecast. One job for the professional might be designing the tools that analyse, interpret and contextualise a system of collective intelligence, rather than selecting and editing content through the prism of news values. Completeness of awareness was explicitly not the goal. The paper also declined to bet on the platform, arguing that what deserved attention were the qualities of micro-blogging — real-time communication, search, link-sharing and the follower structure — because another service could carry them later.

Fifteen years on, the platform has changed its name and most of its rules, and the framing has outlived it: the mechanics of a live stream as an ambient layer over the news is now the ordinary condition of reporting rather than an experiment.