AI, Journalism, and the New Abyss

Will media managers see the future coming this time?

In a nutshell: An “AI in Journalism Futures” report recently released by the Open Society Foundation offers five views of how the current ChatGPT-facilitated newsroom may quickly and radically shift to a new information ecosystem. It doesn’t beat the drum of what “should” happen, rather it opens windows to what plausibly could happen, what to be ready for, and in doing this shows a sophisticated use of the scenario planning method.

Source: https://www.opensocietyfoundations.org/publications/ai-in-journalism-futures-2024

It is said that today’s crisis is no more than yesterday’s failure to plan.

As newsrooms hit multiple crises in the early 2000s, when classifieds went to Craigslist, readers to Facebook, and journalist to the job center, one might justifiably have said that newspaper executives were driving looking in the rear-view mirror. Most failed to adjust their companies adequately or timeously, haemorrhaging shareholder value or entirely collapsing the business.

Now, as artificial intelligence and language learning models cast their long shadow over the journalism industry, there is, at least in some quarters, an effort to raise the quality of understanding of what’s coming, and none better than David Caswell and Shuwei Fang’s  “AI in Journalism Futures” report recently released.

The report, looking 5-15 years ahead, is the fruit of a year of facilitated inputs of almost 1000 people, including journalists, technologists, academics, and civil society advocates. The project was developed and run by the Open Society Foundations, a George Soros-backed NGO.

“AI in Journalism Futures” distinguishes itself in exemplary use of the scenarios format: a structured rendering of different plausible future worlds, presented to challenge the happy management illusion that tomorrow will be like today, only maybe a little faster (the illusion of strategic persistence.)

Here are the scenarios:

1. Machines in the Middle

The “Machines in the Middle” scenario envisions an information ecosystem in which a large portion of journalistic and civic information is gathered, processed, assembled, and distributed via AI. Humans are both the sources of journalistic and civic information, and consumers of this information, but AI mediates nearly every process within the information ecosystem, essentially “becoming the newsroom” for significant portions of the ecosystem. AI can create experiences of journalism and civic information personalized to individual consumers, both in terms of content and the consumption experience itself, including language.

2. Power Flows to Those Who Know Your Needs

The “Power Flows to Those Who Know Your Needs” scenario envisions an information ecosystem in which AI can essentially create any conceivable experience of journalism or information, regardless of format, style, medium, etc., and regardless of the source of information. In this ecosystem in which anything can be produced, the central question becomes what to produce for each individual consumer in every specific consumption situation. Therefore, for an AI-empowered information producer, knowing a consumer’s information needs, wants, interests, and requirements becomes the fundamental gateway to serving that consumer, and thus a source of economic and social power.

3. Omniscience for Me, Noise for You

The “Omniscience for Me, Noise for You” scenario envisions an information ecosystem in which different individuals and different groups in society experience vastly different information realities because of the divergent ways in which they engage with AI. It considers situations in which some people are essentially super-empowered by AI-assisted information tools, while others are essentially cognitively imprisoned by them, as well as pervasive “filter bubbles” caused by algorithmic recommendations that reinforce consumers’ demonstrated interests, and “echo chambers” are distinct information communities that consumers self-select into.

4. AI with Its Own Agency and Power

The “AI with Its Own Agency and Power” scenario envisions an information ecosystem without meaningful human oversight, where AI systems control the gathering and experience of information for most people. It is a situation in which people—consumers, engineers, editors, or executives—gradually give up more and more agency to adaptive AI systems until humans no longer control those systems in any meaningful way. In this scenario these (non-sentient) AI systems become essentially free to direct the flow and experience of information, independent of human oversight or possibly even independent of human understanding.

5. AI on a Leash

The “AI on a Leash” scenario envisions an information ecosystem in which the potential impact of AI on the flow and experience information has been substantially restricted by societies or by the collective action of consumers. In this scenario some of AI’s potential to change or even to improve the information ecosystem is intentionally left hypothetical and unrealized because of concerns about its potential impact on societies, communities, or individuals.

There are a number of things that make this work stand out. First it is non-doctrinaire on what its subject, i.e. journalism, “should be” in the future, and freely considers its dissolution into multifaceted information ecosystems. Only through this kind of un-tethering from the past can we even begin to prepare for the fundamental shifts that are plausible, the prospects of which should be considered and managed now.

Notwithstanding this, the work also does not simply discount the past. The weight of the past (Inayatullah 2023) that is, legacy concepts and social preferences around journalism and its social roles, will continue to act on the future evolution of the industry in every scenario.

The future views begin from an in-common transitional “efficiency phase” that we are now in following the release of ChatGPT. During this phase, AI in newsrooms has been applied primarily to existing tasks, workflows, and products in ways that essentially remain within existing conceptions of the media.

But this transitional phase is a strategic illusion not to be bet on, soon to be replaced by a fundamentally different information ecosystem with new tasks, workflows, products, and business models. These alternative manifestations are what the scenarios envisage. The future is not just the past speeded up. It involves shifts.

The five future views are not presented as mutually exclusive. Aspects of all are expected to be present in whatever compound emerges. What matters in decision-making is to get a broad read on what may plausibly break out from present conditions, and to create readiness for when one or other of these break-outs becomes hegemonic.

The scenarios also avoid collapsing the future-preparedness conversation by putting forward a preferred outcome. That is, mixing up what to be ready for with what we would like to happen.  There is a place and time for advocating future preferences (Gordon, 2020) but when dealing with external environment forces mostly beyond any newspaper or media company or even policymakers control, good decisions are those that act in best alignment with those forces.

Finally, reflecting on the many discussions behind the report, the authors offer this reflection, which is indicative of high-quality understanding of how the future actually gets made and what to expect:

“Throughout the application process and workshop discussions, it became clear that much of the conversation was not actually about AI, nor about journalism, nor about the current or future information ecosystem, but instead about power. It was clear that power, and the potential for transfers of power from one group to another, was the explicit or implicit subject of many of the submitted scenarios as well as the five final scenarios that were distilled from the workshop. Central to these scenarios is a contest for power over who controls AI and toward what ends.”

Refs:

Inayatullah S. (2023) The Futures Triangle: Origins and iterations. https://doi.org/10.1177/1946756723120316

Gordon, A. (2020) Limits and Longevity: A model for scenarios that influence the future. https://doi.org/10.1016/j.techfore.2019.119851