Institutions of Epistemic Vigilance: The Case of the Newspaper Press

ABSTRACT Can people efficiently navigate the modern communication environment, and if yes, how? We hypothesize that in addition to psychological capacities of epistemic vigilance, which evaluate the epistemic value of communicated information, some social institutions have evolved for the same function. Certain newspapers for instance, implement processes, distributed among several experts and tools, whose function is to curate information. We analyze how information curation is done at the institutional level and what challenges it meets. We also investigate what factors favor the cultural evolution of institutions of epistemic vigilance: these include people’s preference for accurate and reliable information and their ability to assess communicated information in view of the source’s epistemic authority; but also contingent historical factors that make it worth – or not – to contribute to the maintenance of institutions of epistemic vigilance. We conclude the paper by considering the challenges and vulnerabilities of these institutions in the Digital Age.


Introduction
We are facing an unprecedented richness of communicated testimonies -a phenomenon called information overload (Eppler and Mengis 2004;Roetzel 2018) -, accompanied by a wide market of cheap testimonial sources (news outlets, social media, etc). This poses a processing problem for information-hungry human agents or informavores (Dennett 1991), endowed only with limited information processing capacities. Which source to trust out of all the potential ones? And which testimonies to engage with? In this paper, we aim to describe what solutions societies have come up with to address this problem, while also explaining potential weak points of these solutions in our current Digital Age.
Compared to industrialized societies, hunter-gatherer groups relied mainly on direct, verbal, face-to-face communications -'all resources, including energy, materials and information, are transferred almost exclusively by direct human-to-human contact' (Hamilton et al. 2007). This had important consequences. In particular, communicating false information came with the costs of being perceived as an unreliable communicator, and thus losing one's influence. This system, occasionally depicted by Quandt as a 'simple society' in terms of its information network structure (Quandt 2012(Quandt , 2011, has changed in major ways in the past 200 years in most societies. Has it changed for the better or for the worse? Does it render us more likely to be misinformed?
One historically important example of an institution of epistemic vigilance is the newspaper press. Obviously, not all newspapers' purpose is to become an institution of epistemic vigilance, but some do fulfil that role, or have done so in the past. The tasks of many newspapers include: (1) Gathering information (reporting); (2) Verifying information (exercising epistemic vigilance using experts and artefacts); (3) Editing information (organizing the verified information into a public representation that is understood by the target audience); (4) Distributing information.
Tasks 2 and 3, if performed, make the newspaper an institution of epistemic vigilance. There are many other such institutions outside of the press. Academia, for instance, maintains, and relies upon, the peer review process. Scientific publication is an institution that consists, among other things, in implementing procedures where manuscripts are reviewed by experts who assess whether it is worth publishing. Arguably, Web Search Engines too implement mechanisms of epistemic vigilance (Heintz 2006).
The existence of institutions of epistemic vigilance raises related empirical questions: 'why did they evolve?' and 'how do they work?' The answer to the why question would identify the factors that made the institution culturally successful. The answer to the 'how' question describes the mechanism that performs the function -checking the epistemic value of communicated information. We treat the why and how questions in turn in the following subsections. Sperber et al. (2010) explain why psychological mechanisms of epistemic vigilance have evolved. It is an argument that pertains to evolutionary biology, with a focus on cognitive biological traits. In order to benefit from communicated information, the argument goes, one must be able to select the information that warrants beneficial belief updating, and select out the information that would mislead us if we were to believe it. Institutions of epistemic vigilance have the same function: enabling people to benefit from communicated information and reduce the risk of being misled. Yet, their evolution is cultural rather than biological (and their implementation is social rather than neuronal). Explaining biological traits as adaptive, and thus as having a function, is a well-grounded method in evolutionary biology. In the social sciences, however, ascribing a function to a social institution must be done with great care. We say that a social institution has a function X, if the institution causally does X and doing so contributes to its cultural stability (Heintz 2007). For instance, individuals might decide to finance the institution doing X because they want X to be done. For institutions of epistemic vigilance, we thus argue that it is what happens when X is replaced by 'exercising epistemic vigilance'. We therefore need to analyze why and in what context people would contribute to the maintenance of institutions of epistemic vigilance. Note that, while the question is one of cultural evolution -why do some ideas and practices spread and stabilise within a community? -we do not presuppose the process to be one of selection and faithful copying (see Heintz and Claidière 2015, for different framework theories of cultural evolution). Rather, the cultural success of institutions might result from some attraction towards some practices. In particular, the production of practices of epistemic vigilance, as well as their selection, are, as we argue below, shaped by psychological mechanism of epistemic vigilance.

Why Do Institutions of Epistemic Vigilance Achieve Cultural Success?
While people do have psychological skills and can exercise epistemic vigilance on their own, there are nonetheless contexts where they can strongly benefit from trusting institutions with epistemic tasks. As shown by Figure 1, Distributing cognitive labor in this way can dramatically increase cognitive efficiency. Indeed, thoroughly assessing communicated information by oneself can be a very demanding task. Two aspects of the current environment make it especially daunting: first, many relevant communicated pieces of information are extremely difficult to assess and yet highly relevant. Think of medical information, for instance. Second, the availability of communicated testimonies have drastically increased in the 19th century and in our Digital age. In such contexts, people can strongly benefit from delegating the task of filtering out irrelevant and finding relevant information, thus decreasing the cognitive cost they would otherwise have to pay. Institutions of epistemic vigilance allow people to economize their own resources, like time and effort (Hames 1992). Thus, by trusting institutions that exercise epistemic vigilance, people efficiently delegate assessment and curation of communicated information. In particular, reliable outlets enable information hungry people to follow a relatively effortless strategy to keep themselves safely and readily informed: they only need to stay vigilant towards the outlet they choose as a source, and need not assess the trustworthiness of the multiple sources from which the information comes from.
In view of the above considerations, our hypothesis is that people are, in certain conditions, willing to delegate tasks of epistemic vigilance to reliable institutions. This willingness constitutes a factor of cultural success for institutions of epistemic vigilance, which perform a function that sufficiently many people recognise as useful. This is to say that some historical contexts, involving the multiplication of sources and other above-mentioned aspects, together with the humans' preference for reliable information, cause cultural evolutionary processes that favor the constitution of institutions of epistemic vigilance. They are, in other words, factors of attraction towards institutions of epistemic vigilance (Sperber 1996;Scott-Phillips, Blancke, and Heintz 2018). In particular, this led US Americans of the late 19th century to invest in journals that were epistemically vigilant. The reliability of institutions of epistemic vigilance can pay off, even though it does not always do. Institutions of epistemic vigilance face several challenges for maintaining their existence, which are, we will argue, well-illustrated with the current crisis sometimes called the 'post-truth era'.

How Institutions of Epistemic Vigilance Work
Let us look at an example of tasks of epistemic vigilance being performed inside a newspaper. Assume that a businessman, Mr. V., had been accused of murdering his business partner. How do we acquire the information that Mr. V. is a suspect, in an ideal world of institutions of epistemic vigilance? First of all, if there was a witness, he spoke to policemen. The police spokesperson communicated the details to a reporter. The reporter brings back what she got to know from the The information processing advantages of the newspaper press. The first scenario to the left, the agent exercises its epistemic vigilance towards a host of different testimonial sources that provide communicated information. With the advent of the institutions of epistemic vigilance, the institution exercises epistemic vigilance on potential testimonial sources as well as the content communicated by them. The institution gathers, curates, then presents a selection of information in a universal format understood by the agent. This means (ideally) that the agent only has to remain vigilant towards the institution, instead of a large number of potential sources.
spokesperson's already filtered account to the press office. If possible, she acquires additional evidence like pictures, interrogation excerpts or security camera footage. She also tries to gather additional witness testimonies. She then shares the draft with her editor. The editor searches for wording mistakes and inconsistencies, evaluates the evidential basis and the newsworthiness of the story. The editor decides whether the journalist had made justifiable conclusions based on available evidence, then sends it to the editor-in-chief for confirmation. The editor-in-chief makes a final decision on whether the article is living up to the standards of the news outlet, then the article is published. Based on personal preferences that were defined by a corporate algorithm, the testimony appears on our screens in a digital format: we are being informed that Mr. V. is accused of murdering his business partner. Consequently, we become averse to do business with Mr. V. in the future, despite the fact that we have not seen the murder, did not gather evidence ourselves, have never talked to Mr. V., the police or any of the witnesses directly.
As the above example shows, institutions of epistemic vigilance distribute the tasks associated with delivering reliable testimonies over several subcomponents. Several people are involved: witnesses, reporters, editors. Artefacts are used such as cameras, sound recorders, and opensource investigative softwares. The tasks are distributed in a systematic way, with each component doing its own share and passing over its output to another component. This systematic organization of cognitive labor forms a distributed cognitive system (Hutchins 1995;Dror and Harnad 2008): it is a cognitive mechanism, since it has a cognitive function, and it is decomposable into elements that implement specified subtasks, following specified procedures. Thus, institutions of epistemic vigilance can be studied as distributed cognitive systems with identifiable components that each have their role within the system. Interestingly, many of its elements will be individuals who exercise their psychological capacities of epistemic vigilance. For instance, a reporter will assess the benevolence and competence of her sources. Yet, institutions of epistemic vigilance are likely to be more than the simple sum of its agents. For instance, one property of these large distributed cognitive systems is the scope of the topics treated that is allowed by the specialization of the reporters.
In this section, we have characterized institutions of epistemic vigilance as social systems that have a cognitive function: to deliver audiences testimonies that are good enough to be the basis of belief-updating. We hypothesized that such institutions have evolved because performing their function did fulfil a need created by information overload, which was increased by social and technological innovations, with the World Wide Web being the last such innovation with a large impact on the informational environment. The need is a consequence of the scarcity of psychological cognitive resources facing the immense task of selecting relevant and reliable information in situations of information overload. Institutions of epistemic vigilance can take this task on because they implement social systems that distribute sub-tasks to sub-components.

The Cultural Evolution of the Press: An Overview
Here we illustrate how an institution of epistemic vigilance evolved using the case of the newspaper press. While institutions of epistemic vigilance can be found outside of the press, we also note that newspapers can function without implementing mechanisms of epistemic vigilance, as this is neither a necessary nor a sufficient condition to become successful. The press provides a good illustration for the points made above, as the traditional press sees its position challenged by the new communication environment.

Disrupting the Communication Environment
The disruption of the communication environment that some have come to call the 'post-truth' age, is not without precedence. The 19 th century rise of mass media is a parallel example to understand the effects of disruption. As it was mentioned before, psychological capacities of epistemic vigilance have biologically evolved to handle verbal communication that mainly took place between people who knew each other. For most of human history, a relatively small number of testimonies arrived through other channels. During the Middle Ages in the West, scarce 'news' had been delivered by clerical networks, university postal services, news singers and couriers (Crombie 2014;Menache 1990). Due to high illiteracy rates, the messages intended for public use had to be proclaimed to audiences. The oral communication environment evolved somewhat slowly, and then changed suddenly during the 19 th century (Slauter 2015;Briggs and Burke 2009). There were nonetheless regions that were outliers to this tendency and had higher literacy rates even before the 19 th century (as well as regions where literacy remained relatively low during the same period). Other factors apart from literacy also limited the emergence of mass informing.
Although newspapers are present from the 17 th century in Europe, a conjunction of factors inside the communication environment was required so they could transform into a medium that bears the main characteristics of a modern newspaper: periodicity, universality, actuality and publicity (Groth 2011;Xavier and Pontes 2019). The cognitive, political and technological limitations were surpassed around the late-19th century in the West.
As mentioned, in most countries the primary cognitive limitation was literacy. During the 19 th century, educational acts were introduced to boost mass literacy. By the end of the century, 97.2% of the male population and 96.8% of the female population had acquired the ability of reading and writing in England, while in the 1841 census data the percentages were only 67.3% among males and 51.1% among females (Harris 1987;Altick 1957;Aspinall 1946).
Next up are political limitations -censorship, monopolies and taxation -on information flow. Due to censorship, available newspapers and other, early forms of public information were often unreliable. Most of the outlets were not allowed to report on domestic politics, only foreign news, and that imposed constraints on their relevance (Pettegree 2014). The interest in newspapers, pamphlets and broadsides increased in times characterized by the lack of centralized control over the flow of information (Raymond 2003). During the 19 th century, strict taxation had been abolished. This was particularly important, as taxation had always been a burden on circulation: it increased prices and put newspapers outside of the reach of working classes (Slauter 2015). Decreasing prices coupled with growing literacy meant that newspapers became both cognitively comprehensible and financially available for large populations. Subsequently, the number of new outlets spiked, and the content of information grew richer. In England, the circulation skyrocketed from 39 million copies to 122 million copies between 1836 and 1854, in the period of decreasing taxes (Lake 1984). Outlets were allowed to cover a wide range of topics, which established universality.
Technology also limited the distribution of information. Printing did exist, but the machinery had been operated manually. During the 19 th century, the steam-powered rotary printing machine and the linotype increased production rates: from around 300 impressions/hour to 20.000 pages/hour (Ward 2015). Locomotives and other means of transportation ensured the periodicity of papers. The telegraph appeared: it allowed newspapers to feature 'breaking' news, ensuring the actuality of papers. The telegraph influenced how information was presented. The straightforward way in which information went through the wires inspired the 'inverted pyramid'-structure that organizes information into a hierarchy of decreasing relevance (Ward 2015;Stensaas 1987). Previously, news had been listed according to the destination of the source rather than the topic or content (Pettegree 2014).
The communication environment had changed rapidly in the West during the 19 th century, in a matter of a few decades. The verbal, personal and infrequent became periodical, written and massdistributed information, for a population that had been largely reliant on direct, face-to-face communication. Aided by the political and technological advances of the 19 th century, newspapers transformed from vessels of propaganda into sources of entertainment for the masses. Subsequently, to handle the hectic communication environment and a never-before-seen overload of testimonies, a number of outlets invested in the position of reliable information curators for the public. Emerging institutions of epistemic vigilance met with success because they solved, for the consumers, the information overload that news outlets themselves have partially created. This new communication environment brought along novel challenges for listeners, new opportunities for deceivers and new tools to construct institutions of epistemic vigilance.

Sensationalism as the First School of Journalism
Providing information for the masses transformed into a booming business during the 19 th century. This did not automatically mean that publishers provided reliable information for readers. Reliability has always been tough work, and tough work is costly. With the circulation increasing rapidly, publishers were out making money by quick return on investments. This resulted in the 'cash-andcarry' format (Holiday 2018; Higdon 2020) of newspaper distribution: issues were sold on street corners by hawkers, in direct competition with other penny papers. A common feature of these 'penny newspapers' were screaming headlines, dramatized drawings and excessive amounts of advertising. Newspapers did not go great lengths to reliably inform audiences about happenings. The aim was to entertain: people bought sensationalist papers as they enjoyed stories about the race of bat people living on the Moon (Griggs 1852) or the bald man who painted a spider on his scalp to scare away flies (MacDougall 1958). The 19 th century Penny Press was notoriously unreliable: this is the era of the Great Moon Hoax and other, historical examples of misinformation.
Journalists of the 19 th century were not yet seen as experts per se, nor had the high social status usually associated with expertise. It did not help much that the practices featured in the early journalism textbooks were often questionable (Phillips 2019). As the ethics of mass informing were still fictile, journalists of the late 19 th century were a mix of sensationalists and professionals. The era was defined by two editors: Joseph Pulitzer and William Randolph Hearst. Building on the legacy of the penny press, the coverage of their 'yellow papers' consisted of overexaggerations of minor events, excessive advertising, faked interviews, pseudoscience, scaremongering and populism (Mott 1941). Hearst himself embodied a different ideal of journalism altogether (Campbell 2006). For him, journalism was not observation, but an active force in shaping the outcome of public incidents. In 1897, Hearst used journalism to directly interfere with another country's domestic politics (Campbell 2002), generating a debate regarding the function of the newspaper. Hearst named his approach 'action journalism' while others condemned it as 'freak journalism'.

Quality and Experts Appear in the Market
The loud and entertaining 'cash-and-carry' remained the main business approach until 1896, when businessman Adolph Ochs invested in The New York Times. In a short Business Announcement (Ochs 1896), Ochs promised to deliver nonpartisan and reliable news to its reader base. 'Without fear or favor' -he wrote. His idea was based on the assumption that if the people were given an actual choice between reliable news and unreliable entertainment, then many would choose the former: We believe that [. . .] thousands would like to buy and read a newspaper of the character and quality of The Times, in preference to [. . .] the papers they have been reading (The New York Times: One Cent! 1898).
Ochs made another important move: he matched the one-penny selling price of his competitors using readership subscriptions.
Ochs' reasoning behind his approach revolved around two interdependent concepts: factuality and impartiality. Without impartiality -ensured by financial independence from political parties and companies -, there was no guarantee of factuality (Ward 2015). These concepts do show a degree of similarity with the two dimensions (see Petty and Brinol 2008;Landrum, Eaves, and Shafto 2015) on which sources are said to be evaluated by epistemic vigilance capacities: competence (with factuality), and benevolence (with impartiality). Essentially, Ochs thought of the audience as consisting of rational agents interested in the state of the world, but trapped in a communication environment that was insensitive to these epistemic preferences. Satisfying other preferences -by scaremongering or entertainment -offered a quicker return on investment for sensationalist papers. The lack of alternative to this type of journalism, however, did not mean that there was no demand for reliability. Ochs was proven right: his sales and circulation increased and other outlets quickly copied his approach (Campbell 2006). Some papers -as their track record of reliable communicators had been established -earned a status that was akin to celebrated brands, like Tiffany's Jewelry or Ford Automobiles. The model introduced by Ochs can be understood as a long-term economic relationship between the reader and the outlet. It allows publishers to build a reputation for themselves as competent and benevolent communicators transmitting quality information, while subscribers have means to give feedback in reader's letters or to outright punish the outlet by terminating subscriptions. The Times often published reader's letters that criticized the editor and the paper itself, which was highly unusual (O'Shea 2008).
To live up to the promises of the Ochsian-model, late 19 th century outlets had to implement novel epistemic practices and signal their reliability and expertise. Publishers needed to make sure that the reporters employed were both competent and benevolent: a demand appeared for experts. Schools were founded to teach journalist competencies. Meanwhile, outlets enacted policies to keep their experts benevolent. Reuters prohibited reporters to be members of a political party. Newsrooms and the organs of fiscal policy became separated; at some news outlets, reporters and managers were not even allowed to travel in the same elevator (Ward 2015). Factual reporting entailed that journalists must leave the newsrooms, interview multiple sources and if possible, provide the names of sources as well as their own. The New York Times itself pioneered technological inventions for news coverage, like photographs and on-the-spot telegram accounts of historic events (Campbell 2006). Institutions of epistemic vigilance that were created around this period slowly became more than the simple aggregation of individual epistemic vigilance capacities. Institutions write down, codify, teach and openly communicate the rules and principles of good epistemic practices in editorial guidelines and newspaper policies. The advantage of the manuals in which institutions describe their epistemically vigilant methods is that they may be updated whenever the changing communication environment requires it. There are also feedback loops, editorial evaluations and other considerations feeding into epistemic practices. Information-curation may evolve quicker than epistemic vigilance, as the psychological capacities are subject to biological, while institutional practices to cultural evolution.
Yellow papers and political propaganda did not vanish after the appearance of the Ochsianmodel, of course, and different means to achieve cultural success coexist. What we see as important is that around the late 19 th century, parts of the press realized that in a turbulent communication environment there was a demand for institutions fulfilling a function analogous to the psychological capacities of epistemic vigilance. The press thus began to develop distributed systems of epistemic vigilance. From a cognitive perspective, this historical process involves groups of the population outsourcing some of the tasks of epistemic vigilance to institutions. This act of delegation -which originates in individuals' epistemic vigilance (the psychological capacities) -contributed to the maintenance of the institutions of epistemic vigilance. Individuals were willing to pay for services provided by institutions of epistemic vigilance; their willingness arose from them being epistemically vigilant, yet confronted with a communication environment they could no longer efficiently curate.

Institutions of Epistemic Vigilance and Digitization
The online version of the 'cash and carry'-type business model is living its renaissance in the Digital Age. Instead of selling copies on street corners, the goal of the employers of this model is to gather as many clicks as possible, for each click and each second spent on a webpage can be monetized. 'Individual messages' (Bak-Coleman et al. 2021) are competing for attention on digital street corners. While some outlets are desperate in pushing towards the subscription-model using Ochs's rationale about impartiality and factuality, these institutions are facing new challenges. It is, with digitization, easier to write any type of information and distribute it widely. It is also possible to communicate anonymously, and thus to get away with lying without having to pay the reputational costs (Acerbi 2020). In this section we highlight four problems that institutions of epistemic vigilance face in the Digital Age.

A Further Increase in the Number of Sources
Digital communication technologies allow virtually anyone to maintain an outlet. The costs of maintaining outlets were cut as printing and manual distribution are no longer necessary. This led to a second explosion of the number of sources of information. The audience, again, has to make a difficult choice regarding which source to believe out of dozens of new ones competing for attention. On the other hand, novel sources entering the market are in competition for a finite number of reputational labels and thus are motivated to attack and to destroy each other's reputation. The attacks manifest in accusations regarding competence and benevolence. Since the traditional cues of reliability and expertise became hackable (see for example, Donovan and Freidberg 2019), it is less evident what strategy should audiences follow to select sources in the digital environment. A conservative information foraging approach that seems reasonable, but often yields suboptimal results, is an increased preference for user generated content, or UGC. UGCs are sometimes perceived as being more benevolent than official sources, since they do not appear to have any vested interests (Liu et al. 2015;Napoli 2016). In reality, UGCs are sometimes created by sources with extreme political views, who are more active in expressing their opinions than moderates, resulting in an overall more polarizing environment (Yldrim, Gal-Or, and Geylani 2013).

Growing Costs of Epistemic Vigilance
News circulation and content curation are no longer simply periodical, like in the 19th and most of the 20th century, when there was a Morning Post, an Evening Standard and a Weekend Special available to get informed. The flow of information from sources is continuous (the so-called 24/7 news cycle), which has unfortunate consequences for journalistic practices. Publication frenzy and time pressures negatively impact journalists and their epistemic practices, as noted by the periodical Worlds of Journalism Study 2012-2016. The mechanisms of information-curation are diluted as journalists face high daily word-counts and gruelling publication requirements (for overviews, see Harro-Loit and Josephi 2020;Nikki 2018;Holiday 2018). This opens up the way for misinformation. Apart from not having time to check sources properly, the publication pressure may push reporters into deliberately faking stories, as several infamous cases illustrate (see Phillips 2019).
The opportunity costs of exercising epistemic vigilance are high in a communication environment that demands new content every minute. In comparison, deceivers are facing reduced financial and social costs for their actions and at the same time, higher benefits. It is less effort to create deceptive or downright fake information than to curate credible sources. The writers of fakes do not have to venture out and talk to witnesses, nor do they have to abide by journalism ethics. Because of this, deceivers work faster. International news agencies sometimes use the expression 'digital gold rush' to refer to the business opportunities of spreading bogus information (Subramanian 2017;Kirby 2016). It would seem that the lack of accountability coupled with the promise of income opportunity favors the creation of cheap and dubious content.

Diminishing Economic Benefits of Epistemic Vigilance
Convincing people to pay for subscriptions is one of the hard problems of Digital Age journalism (Rosenstiel, Jurkowitz, and Ji 2012). When outlets make money through advertising instead of relying on readership subscriptions, they become dependent on actors that may use them as -once againvehicles of propaganda. Reasoning about the epistemic benefits of paying for credible information does not seem to be enough. Ochs himself became successful not only because his paper implemented epistemically sound practices: he also matched the price of his sensationalist competitors. Before the internet, readers had to pay for all types of published information: unreliable or reliable. The contemporary situation is different in that regard. The internet is famous for its norm of freely available information (Moore 2018;Chyi and Lee 2013;Anderson 2009;Himanen 2001;Levy 1986). Meanwhile, good epistemic practices are costly, and the expectation of gratis information forces the institutions of epistemic vigilance to seek funds by monetizing the attention of the reader through advertising or by receiving financial aid from political parties, which in turn damages their impartiality and subsequently, factuality.

Mixing Fake and Truth
Another novel feature of our era is that we use social media platforms to familiarize ourselves with the news, meaning we meet fake and truth in the same epistemic space (Gottfried and Shearer 2016). This epistemic space so far has shown only menial interest to indicate which sources were reliable and which testimony is accurate. Before social media, deceivers had to pay serious costs whenever they aimed to appear on similar platforms as reliable information. Simultaneous presentation with fakes puts reliable information into an uneven competition: truth is insensitive to our preferences, while fakes are doctored to fit them. Since the same epistemic space includes both reliable and unreliable information, the psychological mechanisms of epistemic vigilance -that tend to modulate trust in light of the message source -are decreased in their efficiency.
There are tentative countermeasures by the owners of social platforms that are aiming to regulate the content that appears for users. These often include some kind of a signal that indicates information quality for the customers (Zhang et al. 2018). There are also political pressures pushing for the evolution of institutions of epistemic vigilance within these influential epistemic spaces.
The growing awareness that the communication environment matters raises important moral (and practical) questions regarding, to name the two most important, responsibility of the major actors who shape our communication environment, and exercising freedom of speech on these platforms. Lately, the European Parliament received a proposal known as the Digital Services Act, which would push the legal responsibility of the appearance of illegal content on the platform owner. It would also allow researchers to access the algorithms (which are still mostly opaque), so they could estimate the harm they may or may not generate (Single Market for Digital Services, Proposal 2020/361).

Arms Race between Communicators and Audiences
Mechanisms of epistemic vigilance align the interests of the communicators and the interests of the audience: the audience is interested in paying attention to the communicators because they are less likely to be misled by irrelevant or false information, and communicators are interested in communicating relevant and true information, because otherwise they risk losing the attention of the epistemically vigilant audience (Heintz and Scott-Phillips 2021). We have shown that changes in the communication environment can lead to mechanisms of epistemic vigilance losing their efficiency. Communicators, in turn, are less constrained to provide only relevant and true information. The solution for communication not to collapse, is to update mechanisms of epistemic vigilance so that they catch up in an arms-race between communicators and their audience. The 'Post-truth era' can be considered as the latest chapter in the history of this arms-race. As the 21st century communication environment is disrupted by digitization, communicators are again given new opportunities to deceive and misdirect listeners, and to game institutions of epistemic vigilance in multiple ways. This stage of the arms-race, as we have shown, bears similarities to previous disruptions, as well as features that can be counted as novelties. However, there is no guarantee that new institutions of mechanisms of epistemic vigilance will evolve. The audience might actively seek ways to avoid being deceived and misdirected without social institutions answering this need.

Epistemic Norms for Designing Institutions of Epistemic Vigilance
Mechanisms of epistemic vigilance have evolved so that they perform their function efficiently, which is to assess the epistemic value of communicated information. Efficient mechanisms of epistemic vigilance accurately assess true information as true and false information as false; they do that reliably and at reasonable costs (in terms of cognitive effort, time, money, etc.). In section 2, we characterized institutions of epistemic vigilance as functional institutions and we provided an illustration of how such an institution evolved in the history of the press, and what procedures they implemented. To answer to the challenges specified in the fourth section, however, it would be useful to have criteria about what makes a satisfactory institution of epistemic vigilance that are more detailed than 'it should perform its function'.

Reducing Reliance on Testimonies
Epistemic vigilance crucially deals with testimonies. One philosophical trend has argued that trust in testimonies is rational only if it is grounded in evidence that is acquired without relying on testimony. For instance, I am rational to trust what Thom said to me because I have independent evidence that Thom is trustworthy, and this evidence is devoid of any testimonies. It would be irrational, for instance, to trust Thom on the basis of testimonial evidence from Thom saying he is trustworthy. The structure of this simple example can be found in much more complex forms. For instance, I could trust Thom because Julia tells me he is trustworthy, and trust Julia on that point because Thom told me she is trustworthy. A real-life example is as follows: a Wikipedia article makes the statement that p. This statement is then made in a news outlet on the basis of the Wikipedia article. At the same time, Wikipedia editors request that the statement p be linked to some reference that would provide a guarantee that the statement is not made up (as Wikipedia does implement practices of epistemic vigilance). The Wikipedia article finds a statement that p in a newspaper and includes a reference to it. However, this is the very newspaper that made the statement that p just because it was written in Wikipedia! The testimony that p is thus validated on faulty grounds. In order to avoid such mistakes, the reductionist view of testimonies argues that a rational process for updating beliefs on the basis of testimonies is such that all trust in testimonies is justified on the basis of testimony-free evidence. The mission of mechanisms of epistemic vigilance would then include finding and compiling this testimony-free evidence so as to adequately adjust trust in testimonies (e.g. Hardwig, 1991). (The testimony-free evidence can be about the trustworthiness of the testifier, it need not be on the content of what is testified). By contrast, anti-reductionists have argued that processes that fully reduce reliance on testimony are de facto impossible and, in many cases, not desirable (Coady 1992). For instance, children trust things that they do not understand, which turns out to be an important way to acquire complex yet useful information (Gergely and Csibra 2020;Heintz 2011). In our analysis, we develop a somewhat anti-reductionist view: it is good and efficient for individuals to trust more rather than less; individuals gain from putting their trust in reliable sources of information; possibly, people cannot and should not assess sources and content following some stringent rationalist criteria. Yet reductionist attempts -attempts to form beliefs that are eventually grounded on the basis of testimony-free observations -do have epistemic value. Paradoxically, these attempts are best pursued, in our current environment, collectively. These attempts are best pursued by institutions which distribute cognitive tasks and are maintained by a sufficiently big community of users.
The reductionist/anti-reductionist debate has been focusing on the individual agent faced with the challenge of rationally updating beliefs while drawn in testimonies. In that matter, gullibility has been shown to often be the most rational process, as opposed to always remaining sceptical or to attempting an impossible search for non-testimonial evidence. How does this analysis apply to social institutions of epistemic vigilance? Should such institutions attempt, or not, to find all non-testimonial evidence for or against each testimony? The practice of the press, when it exercises epistemic vigilance, is to use a hierarchy of direct and indirect evidence ranging from photographs (non-testimonial direct) to spokesperson accounts (testimonial indirect). When two pieces of evidence are in conflict, say, the citizen observer's testimony versus a video footage, then non-testimonial evidence is given more weight in the verification process.
Checking the validity of testimonies becomes quite daunting when testimonial chains are long, involve many agents, often cross-reference each other, and unavoidably include contradictions. Institutions of epistemic vigilance, however, need not shy away from the task. Institutions are indeed much better endowed than individuals: they can dedicate more resources to the task, develop expertise, and distribute tasks. In the end, the institutional system might have epistemic skills and competence that completely change the analysis that makes a reductionist program irrational at the individual level.

De-Biasing at the Institutional Level
Institutions include individual human minds. Upon processing information, these minds naturally display various preferences and biases. It would be logical to assume that institutions do not merely mimic some selected capacities of the mind (like epistemic vigilance), but may be prone to mirror the biases that accompany them. Because of this, institutions of epistemic vigilance include de-biasing procedures, for instance by involving several individuals who check each other in an argumentative context (Mercier and Sperber 2017). In addition, the codified editorial guidelines detailing the means of testing information that employees of an institution of epistemic vigilance follow, lead to displaying a level of factuality and impartiality that transcends individual biases. 'The method is objective, not the journalist'-as it was famously put by Kovach and Rosenstiel (2007). Evidence shows that 'forces on the routine level of analysis' -e.g. regulated evaluations of newsworthiness -may be primarily responsible for the quantity of information published on a story about congressional bills (Shoemaker et al. 2001). The personal characteristics of journalists, such as their political ideology, voting preferences or education did not significantly correlate with the quantity of coverage given. Although this evidence can by no means be completely generalized (the original study concentrated on one specific topic), it nonetheless shows that institutions of epistemic vigilance have the potential to attain a better level of impartiality as they operate with multiple individual agents and a codified system of information curation. As for the assessment of testimonies, the fact that institutions of epistemic vigilance are distributed cognitive systems give them desirable epistemic properties that might not be achievable at the individual level.

Using Cognitive Tools of Epistemic Vigilance
The efficiency of distributed cognitive systems can be enhanced with the use of cognitive tools. In the case of epistemic vigilance, weighting algorithms such as PageRank, bots and other programs can be programmed and used for information curation. A more beneficial approach to the threat of misinformation would not be to combat it, but to make reliable information more salient (Acerbi, Altay, and Mercier 2022). In theory, algorithms can be programmed to give space to different opinions and views based on some concept of fairness. If algorithms are responsible for creating and/or fostering opinion-bubbles, then they could also be programmed to burst them (Savage 2019). Scholars and journalists fear that in the near future, AI may be used to produce an endless supply of disinformation (DiResta 2020). If this is true, then a similar AI could be programmed to tirelessly curate reliable information based on the principles of epistemic vigilance. The institutions of epistemic vigilance are challenged due to digitization. Perhaps the solution lies in digitization too: by giving more weight to non-human artefacts inside institutions of epistemic vigilance.

Conclusion: The Epistemic Power of Institutions of Epistemic Vigilance
Trusting is recognisably at the basis of humans' cultural achievements. It figures as a key element of models of cultural evolution (e.g. Henrich and Gil-White 2001). Yet, there are good reasons to think that humans are not simply gullible: they modulate their trust in what is communicated to them considering the trustworthiness of the source (itself dependent on assessments of competence and benevolence), the plausibility of the claim and the reasons in its favor (Sperber et al. 2010;Mercier 2020). We have argued, however, that the evolved psychological mechanisms that modulate trust -the psychological mechanisms of epistemic vigilance -cannot fully account for how people deal with the overwhelming flow of communicated information of our current social environment. A full account of human trusting behavior therefore needs to consider the ways the communication environment is itself shaped: we have argued that it can include institutions that create powerful affordances for psychological mechanisms of epistemic vigilance. These affordances take the form of reputation labels that reliably signal expertise, or simply involve aligning ease of access to the information with relevance and plausibility.
The case of the newspaper press illustrates how the communication environment evolves. There is technology, of course, but also a set of psychological factors that influence what tends to be communicated and how. The study of these psychological factors is at the heart of quite a few researches on misinformation (see Mercier, Majima, and Miton 2018;Mercier and Miton 2019;Altay, Hacquin, and Mercier 2020;Boyer 2018;Mercier 2017). In the case of the press, these psychological factors strongly influence the content of the sensationalist newspaper, but not only. We focused on one set of psychological factors: psychological capacities of epistemic vigilance. These capacities, we argued, constitute factors of attraction for the cultural evolution of institutions of epistemic vigilance.
In spite of its unintuitive aspect, trusting institutions of epistemic vigilance is -we suggest -the most rational way to deal with the wealth of available communicated information. This is because such institutions can be designed so as to perform their function in communication environments where individual psychological capacities are limited, and sometimes outwitted. Institutions of epistemic vigilance have properties that are different from the properties of any individual that participates in it. Therefore, social epistemologists should expand their analysis of trust in other individuals, experts or not, to trust in institutions. If anti-vaxxers are not making the best choice for themselves, it is because they fail to trust those institutions who do, in fact, implement the best mechanisms of epistemic vigilance (in this case, the newspapers that trust medical institutions after checking that they have good reasons to do so, which is that medical institutions have implemented adequate procedures of epistemic vigilance towards scientists). Trusting communicated information that is hard or impossible to check on one's own is an incredible achievement of our societies. It results from the cultural evolution of complex social institutions of epistemic vigilance; the press being a central one, especially for the good functioning of political systems that rely on the assessment of informed citizens.