Introduction

As Hannu Nieminen’s chapter 2 in this volume emphasises, epistemic rights are about knowledge. They are about being informed, understanding the relevance of information, and acting on this information in a way that benefits the individual and society. As with all rights, epistemic rights are fundamentally concerned with equality, with, in this case, equality related to the core inputs for being knowledgeable and informed decision-makers in the democratic process. However, as with so many aspects of economic and political life, the sphere of information is plagued by a wide range of structural inequalities, in which fundamental and established aspects of how the news and information ecosystem is structured and operates systematically undermine the position and opportunities for traditionally marginalised groups.

Building upon this notion of structural inequalities, this chapter zeroes in on what I term here information inequalities (or, in the vocabulary of this volume, epistemic divides)—fundamental characteristics and dynamics of our news and information ecosystem that systematically disadvantage some categories of news and information seekers over others. This chapter seeks to provide an overview of the various categories of information inequalities and to consider how they sometimes intersect, reinforce, and exacerbate one another. This chapter explores this catalogue of information inequalities as sites for potential or ongoing policy intervention and so will also discuss attempted and potential policy interventions to address these information inequalities. In this regard, this chapter operates from the (certainly arguable) position that the acknowledged democratic threats inherent in government interventions into the structure and behaviour of the news and information ecosystem are, to some degree, now being approximated (or perhaps eclipsed) by the democratic threats inherent in government inaction. Many of the examples drawn upon in this chapter come from the U.S. context, though certainly have broader generalisability. That being said, in some specific contexts, international and global examples will be brought to bear as well.

The sections that follow seek to provide a brief overview of a wide range of core information inequalities. Some of the information inequalities discussed in this chapter are long-standing, such as disparities in media ownership, disparate valuations of different audience segments by advertisers (and thus media organisations), and the digital divide. Others are more recent developments, emerging from (or being dramatically exacerbated by) our rapidly evolving digital media ecosystem. These more recently developing information inequalities include the rise of journalism divides, in which some types of communities are provided with much more robust journalism than others; algorithmic bias, which tends to negatively affect some types of digital media users more than others; and the emerging disinformation divide, in which we see certain communities targeted more aggressively with disinformation than others. Of course, as the new information inequalities can compound the effects of the old, the news and information ecosystem becomes that much more incapacitated in terms of equitably serving the needs of the entirety of all people and contributing to the effective functioning of the democratic process.

Advertiser Valuations of Audiences

The fundamental economics of the media and information industries have long contributed to ingrained, inherently structural, information inequalities. We see this foremost in the extent to which the media and information sectors have relied heavily upon advertising revenues for their viability. The needs and interests of advertisers have, therefore, long served as a driving force in the provision of news and information. While the prominence of advertiser support has worked against more fundamental inequalities associated with disparate abilities to pay for news and information, it remains the case that inequalities also arise from the fact that advertisers have traditionally valued some audience segments more than others.

The audience marketplace, as I have written about elsewhere (Napoli, 2003), possesses a number of characteristics that distinguish it from other types of product markets. From the standpoint of information inequalities, the most significant of these is advertisers’ tendency to assign different values to different audience segments. Traditionally, this has taken the form of differential valuations based on demographic characteristics (age, gender, race, etc.). However, as the audience marketplace has evolved, and as media have become more inherently interactive, differential valuations have been able to take form along various behavioural lines (e.g., with more ‘engaged’ audiences worth more; see Napoli, 2011). Certainly, there can be correlations between the demographic and the behavioural dimensions of media audiences. The key point here, however, is that these differential evaluations have ripple effects that impact the nature of the content that advertiser-supported media produce, which can lead to the kind of information inequalities that may merit policy intervention.

For instance, in the early 2000s, the U.S. Federal Communications Commission (FCC) investigated the possibility that certain radio audience segments (notably Black and Latino listeners) were being systematically undervalued by advertisers, due in part to the media and advertising industries operating under inaccurate stereotypes about these categories of listeners (Ofori, 2001). The ramifications of such undervaluation are that content that serves the needs and interests of Black and Hispanic listeners will then be underproduced, creating a fundamental information inequality, in that being a member of a minority community translates into having disproportionately less content produced that addresses your needs and interests or at least less investment in the quality of the content that addresses your needs and interests (Napoli, 2003).

In this case, some tangible policy actions were taken, with the FCC ultimately deciding in 2008 to prohibit broadcast licensees from entering into advertising contracts that included ‘no urban’ or ‘no Hispanic’ dictates—that is, contracts in which advertisers provide explicit direction that Black and/or Hispanic audiences be avoided (note: the term ‘urban’ is often used in the U.S. radio industry in relation to Black listeners, as ‘Urban’ is a recognised radio format label, and refers to programming that traditionally attracts Black listeners).

A more recent manifestation of information inequalities borne of differential advertising valuations can be found in the role that targeted advertising has come to play in the operation of social media platforms. The demographic and behavioural targeting facilitated by social media quickly led to outcomes that attracted the attention of policymakers. For instance, in 2019 Facebook was charged by the U.S. Department of Housing and Urban Development with restricting access to housing-related ads based on criteria such as national origin, familial status, gender, and disability, in violation of the Fair Housing Act (Paul & Rana, 2019). Facebook (since rebranded Meta) ultimately reached a settlement that included disabling certain aspects of its audience targeting functionality within the context of housing advertisements (Feiner, 2022).

In both of these examples, the information inequalities arise from the dynamics of advertiser demand for audiences that are to some extent grounded in persistent prejudice and, to some extent, grounded in the degree to which a correlation unfortunately persists in the U.S. between income and ethnicity. Ultimately, as Hamilton and Morgan (2018, p. 2833) note in their economic analysis of the factors that lead lower-income media audiences to have access to lower-quality news and information, ‘Poor people get poor information, because income inequality generates information inequality. People with low incomes are less likely to be sought out by many advertisers […] This translates into less content meant to aid their decisions or tell their stories’.

Media Ownership

A key long-standing structural inequality in the news and information ecosystem has been the distribution of the ownership of media outlets. Concerns about ownership concentration have long been a defining element of media policy (Napoli, 2001). Within the context of this chapter, the particular concern is the extent to which many traditionally marginalised groups are not even close to proportionately represented within the ranks of owners of media outlets. This pattern can have ripple effects into areas such as content production and employee diversity (Napoli, 2001, 2011). As a recent report from a consortium of public interest organisations noted, ‘People of color comprise roughly 40 percent of the U.S. population, yet remain acutely underrepresented in mainstream newsrooms that, consequently, often under-report or overlook issues of importance to their communities’ (American Economic Liberties Project et al., 2022, p. 1). Such patterns are likely to persist when there is a lack of diversity in the ownership ranks.

Media ownership matters, in terms of providing economic opportunity and opportunities for self-expression, but also in terms of helping bring to bear a greater diversity of ideas and viewpoints, in pursuit of a robust marketplace of ideas (Baker, 2007; Napoli, 2001). From an information inequalities standpoint, those sectors of the population that are not adequately represented amongst the ranks of media owners are not only being disproportionately denied expressive opportunities, but also are less likely to be provided with news and information that addresses their particular needs and interests (Baker, 2007; Napoli, 2001). Moreover, through constrictions on media ownership, the entirety of the media audience is denied access to diverse ideas and viewpoints, creating blind spots that can perpetuate existing biases and help to maintain existing structural inequalities.

We are seeing increased recognition of how these patterns of exclusion in the media ownership realm have had long-term negative repercussions for traditionally marginalised groups. In the wake of the George Floyd murder and the broader conversation about race and structural inequality in the U.S. that emerged in its wake, the news media have initiated self-assessments, with some news outlets acknowledging their failures to serve their communities of colour, a failure that they attribute in part to woefully inadequate diversity in their ownership, management, and staffing ranks (Fannin, 2020; Lowery, 2022; ‘Our Reckoning with Racism’, 2020).

Such actions have been accompanied by calls for ‘media reparations’, which have included, among other proposals, plans for substantive federal investment in Black-owned and targeted media outlets. However, the pattern in the U.S. from a regulation and policy standpoint over the past few decades has been one of scaling back, rather than ramping up, efforts to enhance the diversity of ownership of media outlets. Even the FCC’s (2022, p. 1) recently released strategic plan, which is situated within an explicit concern with gaining ‘a deeper understanding of how the agency’s rules, policies, and programs may promote or inhibit advances in diversity, equity, inclusion, and accessibility’, does not articulate diversification of media ownership as a strategic goal.

Digital Divides

Another category of information inequality arises from disparities across groups in terms of access to information technologies and, relatedly, in terms of the baseline training that different groups bring to the table when provided with access to these technologies. This brings us into the complex world of digital divides, a concept that began with a fairly simplistic conceptualisation (who does—and who does not—have access to the internet) and that has grown more nuanced over time.

The core notion of the digital divide is oriented around the concept of access—specifically, in regard to whether access to the internet (and, later, broadband internet access in particular) is a function of characteristics such as age, ethnicity, income, and geographic orientation (rural v. urban) (Greene, 2021). All of these criteria have been associated with the digital divide, and so, the connection with the notion of information inequalities becomes quite clear. As knowledge-seeking and effective participation in social, economic, and political life have become increasingly tied to the online realm, differential degrees of access across different population segments can have profound implications. This is why achieving equity in internet access has become a core concern amongst many nations around the globe, with the United Nations going so far in 2020 as to declare broadband access a fundamental human right (Salway, 2020).

However, the notion that technology access in and of itself can address the underlying inequities is, of course, a prime example of technological determinism. Moreover, there are nuances within the basic notion of access that can have significant ramifications. For instance, many developing countries (and funders and NGOs working in these countries) have pursued broadband deployment via mobile devices as their primary mechanism for addressing the digital divide. This is a strategy that invites the question—is someone with exclusively mobile device-based internet access on equal footing with someone who has PC or laptop-based access? I have argued in the past that the answer is no (Napoli & Obar, 2014). A mobile device—while certainly providing greater portability—also presents a number of relative constraints (in terms of screen size, keyboard size/ease of use, etc.) that have been found to restrict the range and depth of activities that mobile users engage in relative to PC/laptop users. Such findings raise questions about the possible unintended consequences of policy approaches to addressing the technological aspect of the digital divide that rely on mobile internet access as a substitute for more traditional forms of internet access (Napoli & Obar, 2014).

All of this being said, it is also vital to recognise that any conceptualisation of the digital divide absolutely must extend beyond the degree of access and also take into consideration disparities in the experience and education that individuals bring to the online experience. A substantial body of research has demonstrated that once access is equalised, there remain inequalities in terms of the intensity and manner in which the internet is used across different demographic groups—a phenomenon researchers have termed the second-level digital divide (Hargittai, 2002). Further, researchers have articulated a third-level digital divide, which refers to disparities in the outcomes that arise across different demographic groups, even when controlling for differences in the intensity and manner of usage (van Deursen & Helsper, 2015).

The bottom line is that the same information tool put in the hands of people with different degrees of experience and training can certainly raise the overall baseline, but can also potentially contribute to the expansion of information inequalities, rather than the narrowing. From a public policy standpoint, this means that efforts to combat the technological dimension of the digital divide must be accompanied by educational efforts that can mitigate second- and third-order digital divides.

Journalism Divides

In the U.S. and many other countries, the topic of news deserts has been front and centre in discussions about the future of local journalism (Abernathy, 2020). The notion of news deserts refers to the growing phenomenon in which a community lacks a functioning source of news and information, as a result of the increasingly precarious economics of journalism. In the work that me and my colleagues have conducted on this topic, we employed the terminology journalism divides (borrowing from the digital divide concept discussed above) to more explicitly address the extent to which the robustness of local journalism is a function of the demographic and/or geographic characteristics of individual communities, similar to the patterns we see on the digital divide front (Napoli et al., 2018). So, for instance, our research found that the robustness of local journalism declined as the Hispanic/Latino proportion of the population in a community increased. This is a finding that obviously connects with the audience valuation inequity discussed previously, as the lower advertiser valuations of Hispanic/Latino audiences most likely play a role in undermining the economic viability of local journalism in communities with larger Hispanic/Latino populations. We also found that as the distance to a large media market decreased, so too did the robustness of that community’s local journalism (as it essentially gets strangled by the nearby large-market journalism) (Napoli et al., 2018). Subsequent research has provided further documentation of the extent to which the news desert phenomenon is distributed geographically in ways that raise concerns about information inequalities across categories such as age and income (with older and poorer communities more likely to become news deserts) (Abernathy, 2022). Here again, lower audience valuations in the advertising market likely play an important role in these patterns.

Research has documented a wide range of economic and political harms that befall a community as local news sources evaporate (Hayes & Lawless, 2021; Sullivan, 2020). In this way, the journalism divides category of information inequality can exacerbate other forms of inequality across different community types. And, of course, at the core of this particular information inequality is the extent to which the critical information needs that can often be distinctive to individual communities are being effectively met by local sources of news and information (Friedland et al., 2012).

It is also important to note that such journalism divides do not occur exclusively within the context of geographically defined communities. So, for instance, research has shown that Black Facebook users receive less exposure to accurate and reliable COVID-19-related news and information than other demographic groups (Faife & Kerr, 2021). The exact reasons for this inequity in the dissemination and subsequent exposure to accurate and reliable COVID-19 news and information remain unclear (Faife & Kerr, 2021). The key point here, however, is that the fundamental dynamics that characterise the journalism divides category of information inequality can extend beyond the geographical context of local journalism and also take shape within a context such as ethnic communities on large digital platforms.

From a democratic theory standpoint, accurate news and information is essential to the process of self-governance. The implication here, of course, is that some types of communities (and often those that are already experiencing greater vulnerabilities) are becoming less equipped than others to effectively govern themselves (Usher, 2021).

Disinformation Divides

The digital divide terminology has been further appropriated to inform how we frame current information inequalities within the realm of disinformation. Disinformation has, in many ways, become the signature concern in relation to the contemporary news and information ecosystem (Bernstein, 2021; Wu, 2020). Disinformation has been widely recognised as a global problem (Kaye, 2019), and documenting the prevalence, impact, and digital platform responses to disinformation has become a point of continued focus for both scholars and journalists (Napoli, 2019, 2021).

Research on this front quickly identified an important information inequality—substantially more disinformation was being produced targeting conservative-leaning media users than was being produced targeting liberal-leaning media users, with research (much of it with a U.S. focus) also suggesting that conservative-leaning media users were significantly more susceptible to the intended effects of disinformation than liberal-leaning media users (for a review of this body of literature on ideological asymmetries in disinformation, see Freelon et al., 2020).

When efforts to misinform the population are disproportionately targeted at one sector of the population over the others, we once again find ourselves operating within the framework of an information inequality, particularly when we see that such targeting is leading to disproportional effects across different sectors of the political spectrum. To the extent that conservatives are greater targets of disinformation and more susceptible to disinformation’s effects, as a group they become relatively less equipped to effectively pursue their best interests through the democratic process, operating more as manipulatable political pawns than as autonomous and informed political actors. And to the extent that conservative news and information sources more frequently engage in the dissemination of mis/disinformation than their liberal-leaning counterparts, such outlets are exploiting and exacerbating tendencies already inherent in their audience base.

However, recent research has begun to reveal how the disinformation divide operates along vectors other than the liberal–conservative continuum. Specifically, a growing body of evidence has come to light illustrating how the dissemination, reach, and impact of disinformation are disproportionately affecting communities of colour (Tesi, 2022). Research has found, for instance, that Russia’s Internet Research Agency (IRA) used troll accounts to disproportionately target Black Twitter (since rebranded X) users and that the IRA disinformation targeted at Black Twitter users generated levels of engagement on par with the engagement levels found within disinformation targeted at conservative Twitter users (Freelon et al., 2020).

Concerns about this racially oriented disinformation divide have spurred organised advocacy efforts (Changa, 2021; Lima, 2022), as well as Congressional hearings (A Growing Threat, 2022). Exactly what kind of (if any) policy interventions might materialise to address this disinformation divide remains to be seen. Given the extent to which the U.S.’s strong free speech tradition has extended into the realm of protecting disinformation (Goodyear, 2021), it seems unlikely that we will see meaningful policy responses on this front.

Extrapolating this phenomenon globally, recent reporting derived from the documentation provided by Facebook/Meta whistleblower Frances Haugen has revealed how the company has allocated its content moderation resources across the various countries in which it operates. These allocation patterns reveal enormous inequities that overwhelmingly prioritise and privilege Facebook users in some countries over others (see, e.g., Newton, 2021). For example, Facebook’s own data revealed that company employees collectively spent over 3.2 million hours combating false and misleading information on the platform; however only 13 percent of this time was spent on countries other than the U.S. (Scheck et al., 2021). When we consider that the U.S. accounts for roughly 8 percent of global Facebook users, the fact that the U.S. is the focus of 87 percent of the company’s misinformation mitigation resources provides a powerful sense of another important dimension of the disinformation divide.

Deeper dives into how Facebook has allocated its content moderation resources concluded that ‘many of these markets are in economically disadvantaged parts of the world, afflicted by the kind of ethnic tensions and political violence that are often amplified by social media’ (Simonite, 2021). The platform provides service in many countries in which neither its automated nor human-conducted content moderation systems operate in the countries’ languages (Simonite, 2021).

The fact that platforms such as Facebook are able to launch in countries without having to demonstrate some baseline level of content moderation capacity is just one example of the harmful side effects that have arisen from the absence of a genuinely global system of platform governance. It remains to be seen whether these troubling nation- and linguistic-level information inequalities related to the disinformation divide that were brought to light by Frances Haugen will be addressed by any systemic policy interventions.

Algorithmic Bias

Algorithmically driven automation has become a defining component of our news and information ecosystem (Crawford, 2021; Napoli, 2014). These algorithmic systems have been integrated into virtually every aspect of the production, distribution, and consumption of news and information (Diakopoulos, 2019; Napoli, 2019). However, an expansive and continually growing body of literature has identified a variety of ways in which these algorithmic systems contain ingrained biases that disadvantage certain communities (see, e.g., Benjamin, 2019; Hao, 2022). In many cases, these algorithmic biases are a function of inherently biased data that serve as the decision-making inputs for these algorithms. These biases have also been shown to be a function of inherent biases and blind spots within the coders who construct the algorithms, who very seldom reflect the diversity of the communities that these algorithmic systems serve (Benjamin, 2019; Noble, 2018). These systems can also, unfortunately, learn bias over time via the aggregated behaviours of system end-users.

Algorithmic bias likely plays a role in some of the examples discussed above, such as Black Facebook users receiving lower levels of exposure to authoritative and reliable COVID-19 information (Faife & Kerr, 2021) and various categories of Facebook users not being exposed to housing advertising (Paul & Rana, 2019). Indeed, the biases that can be inherent in the dynamics of audience valuation described above can subsequently find their way into the algorithmic systems that increasingly dictate the placement of online advertising (Blass, 2019). So, for instance, algorithmic bias has been identified as a causal factor in women receiving less exposure than men to STEM-related job opportunities (Lambrecht & Tucker, 2018).

Search engines such as Google have been shown to exhibit racial biases in their search returns; in some cases exhibiting patterns that reflect blatant racism (Noble, 2018). Specific and well-known examples include Google’s image-identification algorithm classifying Black people as gorillas (Simonite, 2018); search returns involving Black and Asian girls containing primarily hypersexualised and pornographic results; and search results for Black teens returning mugshots, with no mugshots amongst the results for a White teens search (Noble, 2018).

This form of information inequality not only affects the ability of historically marginalised groups to meet their information needs; it affects all search users, who are exposed to biased representations of historically marginalised groups. Such dynamics can, of course, further reinforce and exacerbate existing racial biases and prejudices.

Within the context of news and information, we have begun to see some limited regulatory intervention (such as in the housing discrimination context described above). However, more expansive proposals, such as Noble’s (2018) call for the Federal Communications Commission and the Federal Trade Commission to more aggressively police algorithmic bias on search and social media platforms, or calls for more systematic auditing of algorithms to identify potential biases (Napoli, 2019) have yet to gain traction.

Conclusion

This chapter has sought to provide an overview of the various categories of information inequalities that need to be taken into consideration in pursuit of enhancing epistemic rights. This chapter has also sought to discuss each of these information inequalities in relation to policy interventions (or the lack thereof) that have been implemented on their behalf.

This chapter has only scratched the surface in relation to each of these information inequalities and the policy interventions that have been—or could be—implemented on their behalf. Further, this catalogue of information inequalities is not comprehensive; but hopefully it does adequately represent the broad range of information inequalities around which policy interventions can potentially be pursued. In this regard, this chapter can hopefully serve as a jumping off point for deeper and more expansive conversations about the various forms of information inequality that need to be factored into any policy efforts to enhance individual or collective epistemic rights.