Guest Editor’s Note: Seth Lewis
Discovery, Distribution, and Democracy: Key Considerations in the Next Generation of Research on AI and News
A Guest Editor Introduction to the Special Issue
Seth C. Lewis
The hype cycle, developed by the U.S. consulting firm Gartner, famously illustrates a familiar pattern that tends to emerge with successive new technologies or innovations. It begins with a key breakthrough, a “technology trigger.” Then early publicity of high-profile success stories leads to rapidly growing attention for the technology or innovation, culminating in a “peak of inflated expectations.” This is generally followed by a crash in visibility and expectations as the technology or innovation is deemed to have fallen into a “trough of disillusionment,” as interest slackens and many much-hyped experiments fail. However, in time—and this could be in months, years, or decades, or it could never happen at all—a technology’s productive value begins to be recognized anew, with an emergent “slope of enlightenment” that may involve second- and third-generation iterations of the innovation finding some traction in various enterprises. This may finally be followed by a “plateau of productivity” in which the innovation achieves mainstream adoption as its utility becomes more fully and widely apparent. While there are plenty of critiques of the hype cycle—after all, it’s certainly not a scientific analysis of innovation, and it’s hard to tell when a technology is in one phase or another—the framework can serve as a simple heuristic for considering how technologies have developed over time: from the internet (think: 1990s dot-com bubble followed by the Web 2.0 social media era) to the current fascination with the likes of cryptocurrency, NFTs, and the metaverse (around which there has been no shortage of hype lately).
The hype cycle, however imperfect it may be, also serves as a catalyst for considering questions that are pertinent to this special issue: Where do we stand in the development of artificial intelligence (AI)? And what about the application of AI—and associated forms of algorithms, automation, and augmentation—to the context of media, news, and journalism specifically?
Indeed, talk about hype: few technologies in recent decades have gone through as many booms (and busts) in visibility and expectations as we have seen in the oft-inflated hopes and fears associated with artificial intelligence. After an initial burst of interest in machines thinking like humans in the 1950s, there was an “AI winter” for several decades as the promise of such technology failed to materialize—followed by a strong resurgence in recent years fueled by several coalescing trends in tech development: “ever more sophisticated statistical and probabilistic methods; the availability of increasingly large amounts of data; the accessibility of cheap, enormous computational power”; and, with the growing digitalization of many elements of everyday life, “steady progress and cross pollination in these areas [reinvigorating] the feasibility, importance, and scalability of AI” (Cath et al., 2018, quoted in Lewis & Simon, forthcoming). As such, AI now is having something of a moment, both in terms of public awareness and discussion and in productive application. This is true across a vast array of industries and institutions (Mitchell, 2019) as well as, increasingly, in the domain of media and communication (Broussard, Diakopoulos, Guzman, Abebe, Dupagne, & Chuan, 2019; Guzman & Lewis, 2020).
Perhaps some of the hype—or, better put, mystery and magic—of artificial intelligence arises because it’s so often misunderstood and poorly defined, making it hard to compare one person’s invocation of AI with another. In the most general sense, AI refers to the use of computing to assume tasks normally associated with human intelligence. But many fields and industries have their own ideas of what counts as AI and how it works (Boden 2018). So, it is valuable to clarify between General AI and Narrow AI, or “imaginary” AI as opposed to “real” AI (Broussard, 2018). General AI, also called Strong AI, involves a machine possessing intelligence comparable to that of a human—which, for the foreseeable future, remains purely in the realm of science fiction (Mitchell 2019). By comparison, Narrow AI, also called Weak AI, gestures to the more realistic use-cases of AI that we are more familiar with today, often taking the form of machine learning, deep learning, neural networks, and so forth (Boden 2018). This is the training of algorithms to solve carefully bounded problems, and it’s in this latter form of Narrow AI that we find journalistic applications of artificial intelligence.
Journalistic AI, defined as the use of and orientation toward artificial intelligence in newswork (Lin & Lewis, forthcoming), is coming into view as technologies for automating news have become more fully incorporated into the way that news is produced, distributed, and consumed (for a comprehensive overview, see Diakopoulos 2019). This includes a fast-evolving variety of “smart” tools used by reporters and editors, particularly at large and resource-rich news organizations, to gauge sentiment on social media, customize news for tailored audiences, automatically suggest edits on video clips, transcribe interviews at scale, and much more (see examples in Marconi 2020). “The hope,” as one report signaled recently, “is that journalists will be algorithmically turbo-charged, capable of using their human skills in new and more effective ways” (Beckett, 2019).
So, where are we in the hype cycle for AI and journalism? While it is impossible to say for sure, what we can detect is that this crest in enthusiasm surrounding AI has been met by a corresponding swell of scholarly interest from many parts of journalism and communication research. Studies have sought to explain, for example, the potential for AI in investigative reporting (Stray, 2019) as well as its democratically complicated role in adjudicating the news recommendations that people receive via algorithms (Helberger, 2019). Much of this research figures into the wider study of algorithms and automation as defining elements of journalism’s present turn toward increasingly digitized and AI approaches (Thurman, Lewis, & Kunert, 2019). This line of research also ranges, on the one hand, from exploring the depths and details of tools for semi- or fully automated media (Diakopoulos, 2019) to considering, on the other hand, what these developments mean for how we conceptualize—at an ontological level—the essence of what journalism is and how it is communicated (Lewis, Guzman, & Schmidt, 2019). However, it’s also important to acknowledge that research is still at an early stage in capturing the full range of implications—be they sociotechnical, political, economic, ethical, philosophical, legal, and so forth—that arise at the intersection of news, media, and artificial intelligence.
This special issue, therefore, offers a key step forward in advancing this discussion on news in the era of AI. This is particularly so on three fronts: how news is discovered, how news is distributed, and how news figures into broader democratic aims. These 3 D’s—discovery, distribution, and democracy—each point to essential questions that scholars are only beginning to unravel, and which will be crucial to the next phase of research on artificial intelligence and journalism. First, if pattern detection at scale is a key affordance of AI, how could such tools be used by journalists (and others) to surface information that is relevant and newsworthy and thereby expand the capacity and impact of what journalism has to offer? Second, as media consumers increasingly rely on search and social media to encounter news, how are algorithms and related tools of automated media distribution prioritizing certain types of information and thereby shaping the public agenda vis-à-vis the traditional agenda-setting role of journalists? Third, and implicit in the above questions but more spelled out here: What kind of ethical, normative, and ontological implications emerge as journalists embrace automation?
In the first article of this special issue, “Exploring Reporter-Desired Features for an AI-Generated Legislative News Tip Sheet,” the authors—Patrick Howe, Christine Robertson, Lindsay Grace, and Foaad Khosmood—illustrate potential for advancements in what Diakopoulos (2020) calls “computational news discovery.” The authors include a former political reporter, a former legislative chief of staff, a computer scientist, and a scholar of interactive media and communication. “An underlying motivation for the project,” they write, “is to expand journalistic coverage of local governments, in this case state legislatures, by providing a tool for journalists to summarize for them potentially newsworthy developments happening in legislatures.”
Howe and colleagues thus bring a tool-building sensibility and practicality to an important problem: the crisis in coverage of statehouses as local and regional news organizations cut back on reporting staff. They propose AI4Reporters, an “AI system that automatically generates news ‘tip sheets’ generated in response to triggers that are based on analyzing legislative transcripts and other sources for bill information and campaign donations.” The authors try to gauge demand for and use of such a system through a survey of 193 journalists and semi-structured interviews with 10.
The overall response, they say, was positive: a large majority of journalists feel that they don’t have the resources to cover state legislatures adequately, and said they would do more if barriers, such as the lack of ability to track newsworthy events, were removed. Could AI4Reporters help? Respondents expressed enthusiasm for the tool, but they also voiced concerns about trust, transparency, and timeliness. What’s more, the paper also found that journalists want custom solutions, hinting at the challenge of designing a one-size-fits-all solution for improving statehouse coverage. Ultimately, this paper offers an important building block for future study: collaborations among journalists, computer scientists, and related specialists in prioritizing tools that augment the work of news discovery that can be so time-consuming for reporters and editors.
The second article, “Algorithmic Agenda-Setting: The Shape of Search Media During the 2020 US Election,” by Daniel Trielli and Nicholas Diakopoulos, offers a reminder that search algorithms, which often do not receive the level of research attention they deserve in journalism studies, are vitally important to analyze. Such algorithms are key conduits through which people receive and make sense of information in a world where we rely on Google incessantly and where the distribution of news increasingly is disconnected from the original source of its publication.
As we continue to transition from human gatekeeping to algorithmic varieties—and as the place of nonhuman selection systems for news prioritization becomes more prominent moving forward, with the potential for ever-greater personalization—what might we anticipate as a result? This study offers a clue. Trielli and Diakopoulos conducted an algorithm audit of Google to explore how the search engine depicted topics and issues in connection with the 2020 U.S. presidential election. In an extension of the long line of agenda-setting research, the authors ask: “First, to what extent does the distribution of topics selected by search media replicate the agenda of the news media?; and second, to what extent does searcher input alter this distribution?”
The findings, they say, suggest that there are differences between search and news media in the frequency of topics mentioned (e.g., Race and Environment were underrepresented in the news media), and yet they also find that “the ranking of topics—that is, the order of importance in terms of topic prevalence given by search media and news media—is substantially similar.” They also tested whether particular interests by the searcher (by including modifiers in search queries) might rearrange those relative rankings, finding that, no, there is “limited power by the user to reshape the topics in the search results.”
Among other things, this study—combining algorithmic auditing with computational content analysis—itself is an illustration of social science taking advantage of developments in algorithms and automation that are widening the scope of research. This approach serves to illustrate, in this case, that search algorithms may yield variations in agenda-setting, but perhaps not with the level of user-driven influence that may be imagined, altogether indicating the need for more research into the quality, malleability, diversity, and representativeness of search results.
Finally, this special issue concludes with “The Missing Piece: Ethics and the Ontological Boundaries of Automated Journalism,” by Colin Porlezza and Giulia Ferri. This study rounds out our discussion of hype cycles, beginning as it does with a reflection on the overinflated boosterism—bordering on a technology-as-savior mentality—that has accompanied the emergence of algorithms, automation, and AI in journalism. Against that backdrop, and with a thorough review of the academic literature to date, the authors seek to tease out ethical and ontological dimensions that they believe have particular resonance for the democratic role of the news media and how that role may evolve in an AI era. “A profit orientation that accompanies the implementation of automation,” they write, “is not in itself a problem given that news organizations are driven by profits, but if taken as the main ontological reason behind the use of AI in journalism, it can have a fallout that transcends the boundaries of newsrooms.”
As part of a larger study of perceptions about journalism innovation, the authors conducted qualitative interviews in five countries (Austria, Germany, Spain, Switzerland, United Kingdom). In each country, some 20 experts from the news industry (journalists, editors, and media managers) and from academia (journalism scholars) were asked about the 20 most important journalistic innovations in the previous decade. Overall, more than a thousand innovations were cataloged through their research; 56 of these were related to news automation or AI. By analyzing if and how interview respondents legitimized these innovations in news automation, the researchers focused on what their perspectives indicated about evolving ideas about the ethics and ontology of journalism (e.g., what is journalism and what is it for?).
“The results show that automation is often viewed through an economic lens,” they note, “offering opportunities to increase the efficiency of news production, personalization, or increase time for more complex investigations”—but without a corresponding concern for questions about what constitutes journalism, what its purpose is, and how it is to be distinguished from other information products. Their findings point to the need for future study into how AI not only changes how journalism is done but why, pointing to essential normative questions about journalism and democracy that deserve further scrutiny in connection with artificial intelligence and the opportunities and challenges it portends (see Lin & Lewis, forthcoming).
In conclusion, regardless of whether the hype cycle offers a lens into the ebbs and flows of expectations for AI and journalism, what we can say conclusively is that as shifts in news discovery and distribution change in accordance with algorithms, automation, and augmentation, we need research to chart and critique those developments, and to illustrate their implications for the relationship between news and democracy. More broadly, we need scholarship that helps industry and academia alike to understand how shifts in relations between humans and machines may help or hinder the well-functioning of journalism, including the ultimate role that it can play in cultivating conditions for the good life.
Beckett, C. (2019). New powers, new responsibilities. A global survey of journalism and artificial intelligence. Polis. London School of Economics. https://blogs.lse.ac.uk/polis/2019/11/18/new-powers-new-responsibilities/
Boden, M. A. (2018). Artificial Intelligence. A Very Short Introduction. Oxford University Press.
Broussard, M. (2018). Artificial Unintelligence: How Computers Misunderstand the World. MIT Press.
Broussard, M., Diakopoulos, N., Guzman, A.L., Abebe, R., Dupagne, M., & Chuan, C. (2019). Artificial Intelligence and Journalism. Journalism & Mass Communication Quarterly 96 (3): 673–695. DOI:10.1177/1077699019859901
Cath, C., Wachter, S., Mittelstadt, B., Taddeo, M., & Floridi, L. (2018). Artificial Intelligence and the ‘Good Society’: The US, EU, and UK approach. Science and Engineering Ethics, 24(2), 505–528. https://doi.org/10.1007/s11948-017-9901-7
Diakopoulos, N. (2019). Automating the News: How Algorithms Are Rewriting the Media. Harvard University Press.
Diakopoulos, N. (2020). Computational news discovery: Towards design considerations for editorial orientation algorithms in journalism. Digital Journalism, 8(7), 945–967.
Guzman, A.L. & Lewis, S. C. (2020). Artificial Intelligence and Communication: A Human–Machine Communication Research Agenda. New Media & Society 22 (1): 70–86. DOI:10.1177/1461444819858691
Helberger, N. (2019). On the Democratic Role of News Recommenders. Digital Journalism, 7(8), 993–1012.
Lewis, S. C., Guzman, A. L., & Schmidt, T. R. (2019). Automation, Journalism, and Human–Machine Communication: Rethinking Roles and Relationships of Humans and Machines in News. Digital Journalism, 7(4), 409–427. https://doi.org/10.1080/21670811.2019.1577147
Lewis, S. C., & Simon, F. M. (forthcoming). Why Human-Machine Communication Matters for the Study of Artificial Intelligence in Journalism. In Andrea L. Guzman, Rhonda McEwen, and Steve Jones (Eds.), The SAGE Handbook of Human-Machine Communication. SAGE Publishing.
Lin, B., & Lewis, S. C. (forthcoming). The One Thing Journalistic AI Just Might Do for Democracy. Digital Journalism.
Marconi, F. (2020). Newsmakers: Artificial Intelligence and the Future of Journalism. Columbia University Press.
Mitchell, M. (2019). Artificial Intelligence: A Guide for Thinking Humans. Pelican.
Stray, J. (2019). Making artificial intelligence work for investigative journalism. Digital Journalism, 7(8), 1076-1097.
Thurman, N, Lewis, S. C., & Kunert, J. (Eds.) (2021). Algorithms, Automation, and News: New Directions in the Study of Computation and Journalism. Routledge.