Modern communication systems are often justified as devices for widening access to information, accelerating the circulation of ideas, enabling social contacts, and enlarging the range of voices that can enter public debate. Yet greater informational abundance does not necessarily strengthen public judgment. Under certain political-economic conditions, it can instead weaken the social processes through which claims are assessed, contested, and revised. While information systems have promised to expand human understanding and knowledge, scholars of propaganda, media power, and ignorance production have long shown that they can also facilitate misunderstanding, manipulation, and the strategic organization of not knowing. The central puzzle addressed in this article is therefore not whether societies possess more information than before but why expanding informational capacity can coincide with declining capacities for collective reasoning.
To capture this paradox, the article proposes the concept of opinionrot: the gradual erosion of collective reasoning capacities as cognitive biases become structurally embedded and exploited within information infrastructures. This article understands “opinion” as publicly circulating judgments about shared affairs, and “reasoning” as the collective capacity to justify, contest, and revise those judgments. Opinionrot names the degradation not simply of information quality, but of the conditions under which public opinion can become publicly reasoned. Under conditions of opinionrot, opinions do not disappear. Rather, they become increasingly detached from the deliberative processes through which they might otherwise be challenged, modified, or collectively validated.
The concept addresses a contemporary transformation that extends beyond familiar concerns about misinformation, propaganda, or conspiracy thinking. Such phenomena long predate digital media. What has changed are the structural conditions governing the production, circulation, validation, and visibility of information. Communication infrastructures no longer merely transmit content; they increasingly sort, rank, predict, and monetize attention, often privileging speed, salience, repetition, and engagement over scrutiny and reflection (Chun Reference Chun2021; Seaver Reference Seaver2019). As a result, public reasoning becomes progressively shaped by systems optimized not for deliberation but for prediction, behavioral influence, and data extraction.
This development is urgent in an era in which predictive algorithms not only distribute opinions but also increasingly participate in their production. From McLuhan’s ([1964] Reference McLuhan1994) observation that “the medium is the message” to today’s algorithmic infrastructures of attention, communication technologies have never simply carried information; they have also structured the conditions under which information becomes meaningful. Today, generative artificial intelligence (AI) systems introduce a further shift by enabling the large-scale production of synthetic communicative environments: algorithmically generated texts, recommendations, summaries, and opinion signals that may resemble public discourse while lacking many of the institutional and interpersonal conditions that traditionally supported public reasoning, accountability, and source transparency.
Democratic publics require communicative spaces in which claims can be traced, contested, and revised under conditions of at least partial communicative equality. Following Arendt (Reference Arendt1961), opinion is not a private belief but a public practice of judgment that depends on plurality, exchange, and contestation. Habermas (Reference Habermas, Burger and Lawrence1989) similarly conceives of the public sphere as the communicative infrastructure through which such judgments circulate and acquire legitimacy. Opinionrot matters because it helps to explain why these conditions become increasingly difficult to sustain, even when they are not entirely extinguished. The argument is therefore not that rational public judgment has disappeared, but that the infrastructures supporting it are subject to mounting structural pressures. Pockets of collective reasoning undoubtedly persist; indeed, their persistence highlights by contrast the conditions whose erosion this concept seeks to explain.
The account developed here focuses primarily on digitally saturated, market-driven public spheres in liberal democracies. It does not claim universal applicability across all societies or historical periods. Rather, it examines how particular configurations of platform capitalism, algorithmic mediation, and AI-driven communication have intensified longer-standing tendencies in the political economy of information. These developments are inseparable from the concentration of informational power within a small number of corporate infrastructures that increasingly mediate public attention while remaining only weakly subject to democratic oversight (Pasquale Reference Pasquale2015; Zuboff Reference Zuboff2019).
Although the current phase of digital transformation gives opinionrot new visibility, the phenomenon has deeper historical roots. Early traces can be found in the media’s role in organizing state power and surveillance (Innis Reference Innis1950), propaganda’s management of belief (Ellul Reference Ellul1957), the political economy of mass media (Herman and Chomsky [2004] Reference Herman, Chomsky and Chorbajian2021), and the conversion of communication into capital (Dean Reference Dean2005). What distinguishes the present is not the existence of manipulation or cognitive bias, but the degree to which such tendencies have become programmable, scalable, and embedded within communication infrastructures themselves.
The argument reconstructs this development through three broad phases of digitalization. The first phase (roughly 2000–10) involved the externalization and delegation of memory to search engines and digital archives. The second phase (2010–20) intensified informational overload and fragmentation through social media, continuous attention capture, and accelerated news cycles. The third phase (2020–present) is characterized by increasingly opaque forms of algorithmic integration in which AI systems anticipate, curate, and prestructure interpretation before conscious judgment occurs. Each phase represents a further institutionalization of cognitive bias within information infrastructures and a corresponding weakening of the social conditions required for collective reasoning.
This essay serves as a first conceptual stepping stone toward a theory of opinionrot. Its ambition is not to offer a full causal model, but to provide an analytic vocabulary and historical frame for understanding how reasoning capacities decay with increasing automation. Methodologically, the article proceeds through a conceptual-historical reconstruction. Rather than predicting future developments or empirically testing causal claims, it traces how successive configurations of persuasion, mediation, and prediction gradually transformed knowledge from a matter of public judgment into an object of probabilistic management. By mapping these conditions of possibility, the article offers a framework for understanding not merely that collective reasoning is under strain, but how its erosion has been historically produced, technologically mediated, and economically incentivized.
What Opinionrot Is
Communication systems do not simply expand what can be known; they also reorganize the conditions under which claims become visible, credible, and actionable. From print publics, to broadcast networks, to social platforms and AI interfaces, systems of mediation have continually reshaped what counts as credible knowledge and who is authorized to speak. The cumulative effect of these transformations is not merely acceleration or amplification, but gradual reconfiguration of the infrastructures through which public judgment is formed.
The metaphor of “rot” is intended to capture that process more precisely than terms such as crisis, collapse, or disruption. Crisis implies rupture. Rot denotes accumulation, sedimentation, and normalization. It describes how communicative environments progressively absorb and institutionalize tendencies toward sensationalism, polarization, cognitive automation, and behavioral manipulation. In this respect, the concept overlaps with research on disinformation and information disorder, which demonstrates how economic and algorithmic incentives often reward affect over accuracy and engagement over deliberation (Pennycook and Rand Reference Pennycook and Rand2021; Wardle and Derakhshan Reference Wardle and Derakhshan2017). Yet opinionrot shifts attention from problematic content to the infrastructures that make such content systematically profitable and scalable.
Opinionrot emerges when communication systems increasingly optimize for persuasion, attention capture, and prediction rather than collective reasoning. Across successive phases of media development, cognitive biases become progressively embedded within information infrastructures themselves. What begins as isolated techniques of influence gradually becomes a structural feature of public communication.
Persuasion: From Public Reason to Opinion Management
The first layer of opinionrot emerged when persuasion became institutionalized as a technique of governing mass publics. What had long been a matter of rhetoric and argumentation was transformed during the twentieth century into a system for managing attention, belief, and consent at scale.
In Public Opinion, Lippmann ([1922] Reference Lippmann1992) describes a world in which citizens could not possibly apprehend the complexity of modern life directly; they navigated instead through images constructed and circulated by the press. Bernays (Reference Bernays1928, 37), in Propaganda, argues that democratic societies required the “conscious and intelligent manipulation of the organized habits and opinions of the masses.” Ellul (Reference Ellul, Kellen and Lerner1965) would later generalize this logic in Propaganda: The Formation of Men’s Attitudes, showing that propaganda is not simply a tool of authoritarian control but an ambient condition of technologically mediated societies for maintaining social coherence under conditions of informational overload.
The significance of these developments is not simply that persuasion became more effective. Rather, persuasion became infrastructural. Public relations departments, advertising agencies, and state communication offices transformed influence into a routinized and professionalized activity. Legitimacy increasingly depended on managing perceptions, shaping plausibility structures, and stabilizing collective moods.
Many of these techniques relied upon cognitive tendencies that would later become central to behavioral science and algorithmic recommendation systems. Bernays’s campaigns exploited forms of social proof and availability; advertising industries refined techniques of affective priming and identity reinforcement; Lippmann’s “pictures in our heads” anticipated contemporary concerns about mediated reality construction. Long before machine learning quantified cognitive bias, modern persuasion industries were already operationalizing it.
Attention: Communication as an Economic Resource
If persuasion established the management of belief, the media-industrial and digital eras transformed attention into a commodity. Mass print, radio, and television centralized communicative power and scaled influence to unprecedented levels. Herman and Chomsky’s ([2004] Reference Herman, Chomsky and Chorbajian2021) propaganda model demonstrates how structural filters systematically shape news production within capitalist democracies. Smythe ([1981] Reference Smythe, Durham and Kellner2006) further argues that advertising-supported media do not primarily sell content but audiences. McCombs and Shaw (Reference McCombs and Shaw1972) show how agenda setting influences what citizens think about, while White (Reference White1950) demonstrates how gatekeeping practices filter social reality before it reaches public debate. As Bourdieu (Reference Bourdieu and Ferguson1998) observes, competition for visibility has increasingly compressed public discourse into spectacle.
The rise of digital platforms intensified rather than eliminated these dynamics. Early internet imaginaries promised decentralization, participation, and democratic renewal (Benkler Reference Benkler2006; Castells Reference Castells2000; Papacharissi Reference Papacharissi2002). Events such as the Arab Spring appeared to validate these hopes (Ghonim Reference Ghonim2012). Yet scholars of communicative capitalism argued that networked participation was rapidly absorbed into new forms of economic extraction. Dean (Reference Dean2005) shows how circulation increasingly substituted for deliberation, while Andrejevic (Reference Andrejevic2009) demonstrates how user participation itself became a source of value generation.
What changed was not simply who communicated, but how communication was valued. Visibility, engagement, and circulation increasingly became ends in themselves. Public discourse was reorganized around metrics of attention rather than standards of justification.
Prediction: From Cognitive Bias to Infrastructural Epistemics
The contemporary phase of opinionrot emerges when communication systems move beyond managing attention to predicting and shaping behavior. The pathologies commonly associated with digital media (filter bubbles, echo chambers, polarization, doomscrolling, news avoidance, and information overload) are better understood as symptoms of this broader transformation.
Research increasingly demonstrates that platform architectures shape not only opinion formation but also affective and cognitive life itself. Studies document compulsive platform use, anxiety, depression, shrinking attention spans, reduced cognitive control, disrupted sleep, social comparison, loneliness, and growing patterns of news fatigue and avoidance (Firth et al. Reference Firth, Torous, Stubbs, Firth, Steiner, Smith and lvarez‐Jimenez2019; Jiang and Ngien Reference Jiang and Ngien2020; Scott, Biello, and Woods Reference Scott, Biello and Woods2019; Skovsgaard and Andersen Reference Skovsgaard and Andersen2020; Twenge, Martin, and Campbell Reference Twenge, Martin and Campbell2018; Wang et al. Reference Wang, Frison, Eggermont and Vandenbosch2018).
These developments are significant because they reveal a shift from individual cognition to infrastructural epistemics. What appears as personal dysfunction often reflects the systematic organization of attention and affect around commercial imperatives. Attention is no longer merely a psychological capacity but a political-economic resource governed by platforms, interfaces, and metrics (Chun Reference Chun2021; Pedersen, Albris, and Seaver Reference Pedersen, Albris and Seaver2021; Seaver Reference Seaver2019).
A central mechanism is the exploitation of negativity. Research across psychology, political behavior, and behavioral economics demonstrates that negative information disproportionately shapes attention, memory, and judgment (Kahneman and Tversky [1979] Reference Kahneman, Tversky, MacLean and Ziemba2013; Lau Reference Lau1985; Nannestad and Paldam Reference Nannestad and Paldam1997; Soroka Reference Soroka2006; Soroka, Young, and Balmas Reference Soroka, Young and Balmas2015; Vonk Reference Vonk1996). Behavioral economics describes this asymmetry as loss aversion (Kahneman and Tversky [1979] Reference Kahneman, Tversky, MacLean and Ziemba2013). The familiar journalistic maxim “If it bleeds, it leads” captures a long-recognized tendency within news production (Glassner Reference Glassner1999; Kovach and Rosenthal Reference Kovach and Rosenthal2001).
Opinionrot does not arise because humans are biased. It arises because contemporary information infrastructures can learn, quantify, and systematically exploit those biases. Algorithmic systems detect, rank, test, and amplify emotionally salient signals in real time, making outrage, fear, and perceived decline especially profitable forms of communication (Müller et al. Reference Müller, Schemer, Wettstein, Schulz, Wirz, Engesser and Wirth2017; Tufekci Reference Tufekci2015). In competitive platform environments, algorithms are optimized to maximize engagement and retain user attention (Bolaño Reference Bolaño2015).
This transformation can be understood through dual-process theories of cognition. Kahneman (Reference Kahneman2011) distinguishes between system one reasoning (fast, automatic, associative, and affective) and system two reasoning (slow, reflective, and analytical). Opinionrot can be understood as the large-scale infrastructural capture of system one. Digital platforms externalize, automate, and monetize intuitive cognition while weakening the conditions that sustain reflection, deliberation, and epistemic friction (Mercier and Sperber Reference Mercier and Sperber2017; Pennycook and Rand Reference Pennycook and Rand2021).
The result is an increasingly accelerated information environment in which the tempo of circulation exceeds the tempo of judgment. Individuals confront a near-unlimited supply of information but diminishing opportunities for careful evaluation. Under such conditions, abundance often produces fatigue rather than understanding. Algorithmically curated news feeds privilege engagement over diversity (Thurman and Schifferes Reference Thurman, Schifferes and Franklin2015); synthetic reviews and automated commentary generate manufactured consensus (Keller and Kleinen-von Königslöw Reference Keller and Königslöw2018); and the proliferation of pseudo-public speech contributes to epistemic exhaustion and declining motivation to distinguish between credible and inauthentic information (Lewandowsky, Ecker, and Cook Reference Lewandowsky, Ecker and Cook2017).
Opinionrot therefore names not a crisis of information but a transformation of the conditions under which collective reasoning occurs. The historical movement from persuasion, to attention capture, to prediction reveals how communication infrastructures increasingly absorb cognitive bias into their normal operation. What decays is not opinion itself, but the capacity to subject opinion to sustained processes of public reasoning, contestation, and revision.
Historical Conditions of Possibility
The preceding section argued that opinionrot is not reducible to misinformation, irrationality, or declining attention alone. It names a deeper transformation: the gradual automation of collective reasoning failures. Cognitive biases are no longer merely exploited by propagandists, advertisers, or media institutions; they are learned, modeled, and reinforced through algorithmic feedback. What earlier appeared as communicative capitalism becomes, in the present phase, a precondition for bias automation.
This section identifies the infrastructural conditions that make opinionrot possible. Rather than offering a chronological genealogy, it organizes the argument analytically around three interlocking mechanisms: infrastructural decay, synthetic saturation, and automated public sense making. These mechanisms correspond to what recent debates have described as enshittification, slopification, and murmuring. Together, they explain how digital systems do not merely circulate distorted judgments but increasingly build those distortions into the ordinary operation of search, ranking, recommendation, and generation.
Infrastructural Decay: Enshittification and Bias Concentration
Doctorow’s (Reference Doctorow2025) account of enshittification offers a succinct political-economic account of platform decay. Platforms initially optimize for users, then for business customers such as advertisers or sellers, and finally for owners through increasingly extractive profit-maximizing systems. Seen through the lens of opinionrot, this is not only a decline in service quality; it is also a degradation of the communicative conditions under which public judgment is formed.
As Zuboff (Reference Zuboff2019) argues, surveillance capitalism thrives on behavioral modification rather than communicative truth. This is especially consequential when communicative infrastructures are concentrated in the hands of a small number of owners capable of shaping visibility, agendas, and amplification at scale. The issue is therefore not simply that platforms host bad content, but that they commodify the very conditions of noticing, dwelling, and responding (Pedersen, Albris, and Seaver Reference Pedersen, Albris and Seaver2021).
This process can be understood as bias concentration. Recommendation, ranking, and visibility systems reward engagement, outrage, and repetition because these forms of attention are predictable and monetizable (Tufekci Reference Tufekci2015). Platforms translate cognitive predispositions into data assets and behavioral patterns into revenue streams. Individual-level biases such as negativity bias and confirmation bias are thus scaled through infrastructural incentives. The platform does not merely amplify bias; it learns which biases are profitable, encodes them, and upscales them as engagement fuel.
Chun’s (Reference Chun2021) account of homophily sharpens this point. Recommender systems do not simply mirror preexisting similarity or polarization; they actively sort users into correlated neighborhoods of sameness. Platform decay is therefore also epistemic decay: infrastructures become better at producing predictable clustering than at sustaining plural, contestable, and revisable judgment.
Synthetic Saturation: Slopification and the Collapse of Signal
If enshittification names the political economy of platform decay, slopification names its communicative ecology. The term “slopification,” popularized in technology journalism, refers to the proliferation of low-quality, machine-generated content across digital environments (Mahdawi Reference Mahdawi2025; Marichal Reference Marichal2025; Read Reference Read2024). Slop includes text, images, audio, and video generated at scale and optimized for engagement rather than epistemic accuracy, aesthetic value, or public usefulness.
Empirical evidence already documents the rapid proliferation of AI-generated content across news, search, and social platforms (Fletcher and Nielsen Reference Fletcher and Nielsen2024; Jiang et al. Reference Jiang, Guo, Hu and Gong2024). The economic logic is straightforward: synthetic content is cheap to produce, infinitely scalable, and well aligned with attention economies. Under these conditions, informational abundance becomes difficult to distinguish from epistemic exhaustion. Users confront a deluge of plausible but low-fidelity material, intensifying information disorder (Wardle and Derakhshan Reference Wardle and Derakhshan2017) and epistemic fatigue (Peters and Dunwoody Reference Peters and Dunwoody2016).
Slopification erodes the signal-to-noise ratio of the public sphere. It undermines trust in legitimate sources and exhausts the cognitive resources required for critical evaluation. Even credible claims must now compete within environments saturated by cheap, fluent, and endlessly reproducible approximations. In this sense, slopification is not simply a problem of low-quality content; it also destabilizes the relationship between authorship, authority, and circulation.
Automated Public Sense Making: Murmuring without Publics
Durrheim and Quayle’s (Reference Durrheim and Quayle2025) concept of murmuring describes the distributed, low-intensity exchanges through which publics generate meaning. Murmuring refers to the everyday circulation of sentiments, impressions, and micro-opinions that collectively shape the background texture of social knowledge. People collect fragments of information and use them to infer the social structure underlying information flows. Information thus becomes a signal of social sorting, helping individuals to simplify and interpret the social world (Morselli et al. Reference Morselli, Durrheim, Beramendi, Sepahpour-Fard, Mafunda, Torres, Ohamadike, Quayle and Schuld2025).
This process is not formal deliberation, but it matters for public judgment. It resembles Noelle-Neumann’s (Reference Noelle-Neumann1974) “spiral of silence” in motion: people infer what others believe, what can be said, and which claims are publicly viable. Collective judgment emerges not only through explicit debate, but also through ambient social sensing.
In the generative era, this murmuring can become automated. Large language models integrated into search engines and public-facing interfaces ingest vast corpora of human utterances, identify statistically salient patterns, and reproduce the appearance of collective reasoning without its underlying plurality. They generate a synthetic semblance of public opinion: a probabilistic echo of prior discourse that may appear consensual, authoritative, or socially grounded while lacking reciprocity, accountability, and situated exchange.
This is why recent work on language and AI is crucial. Kockelman (Reference Kockelman2024) shows that large language models do not simply transmit discourse more efficiently; they generate discourse-like outputs by modeling statistical regularities across prior utterances. Weatherby and Justie (Reference Weatherby and Justie2022) similarly help to clarify how such systems can simulate socially grounded language without participating in the accountable and indexical practices through which public judgment ordinarily becomes meaningful. What is produced is not deliberation, but a fluent approximation of deliberation’s surface features.
The psychological consequences are significant. Automated murmuring can amplify affective contagion and feedback loops of emotional resonance, potentially intensifying polarization, extremism, radicalization, and other forms of social disintegration (Cho et al. Reference Cho, Ahmed, Hilbert, Liu and Luu2020). Opinionrot reaches a new stage here: not only are judgments ranked and nudged, but the background through which people sense public opinion is also increasingly generated in advance by predictive systems.
Convergence: From Post-Truth to Probabilistic Truth
Enshittification, slopification, and automated murmuring converge in a broader post-truth condition. Post-truth should not be understood as a historical rupture between an earlier age of truth and a later age of falsehood. Rather, it names a mutation in the sociotechnical conditions under which truth claims acquire credibility: from editorial authority and slow gatekeeping to algorithmic ranking, engagement metrics, multimodal persuasion, and predictive coherence.
In this environment, truth no longer travels primarily from facts to statements through verification. Increasingly, credibility is shaped by fluency, coherence, visibility, and predictive fit, as exemplified by the outputs of large language models. The transition from truth verification to pattern recognition marks a fundamental epistemic shift. Whereas classical knowledge systems relied on the validation of sources, large-scale AI systems operate through the prediction of linguistic or behavioral continuities (Bender et al. Reference Bender, Braun, Möller and Gehrke2024; Floridi Reference Floridi, Miller and Taddeo2020). Large language models function as probability engines of belief: they do not assess veracity but compute coherence within vast statistical spaces. Their authority often derives less from evidence than from fluency (Birhane Reference Birhane2021).
In McLuhan’s ([1964] Reference McLuhan1994) terms, the medium (now algorithmic infrastructure) has once again become the message: the process of generation substitutes for the process of justification. The epistemic contract between statement and world is rewritten as a probabilistic relation between token and dataset. When information is generated through probabilistic inference rather than verified reference, trust, deliberation, and political judgment become structurally destabilized (Durrheim and Quayle Reference Durrheim and Quayle2025).
This drift is not only epistemological but also civic. When digital media infrastructures and AI systems shape attention, emotion, and discourse at scale, they alter the conditions of democratic reasoning itself (Helbing et al. Reference Helbing, Frey, Gigerenzer, Hafen, Hagner, Hofstetter, van den Hoven, Zicari, Zwitter and Helbing2018). Empirical indicators of this drift are increasingly visible: declining diversity in news and search results (Fletcher and Nielsen Reference Fletcher and Nielsen2018), linguistic homogenization across digital discourse (Sourati et al. Reference Sourati, Karimi-Malekabadi, Ozcan, McDaniel, Ziabari, Trager, Tak, Chen, Morstatter and Dehghani2025), and the expansion of AI-driven content for information seeking (Chowdhury and Chowdhury Reference Chowdhury and Chowdhury2026).
The result is a shift from agonistic pluralism toward algorithmic consensus, where disagreement risks being smoothed into statistical averages. If public reasoning depends on exposure to diverse standpoints (Mouffe Reference Mouffe2013), the collapse of disagreement into modeled coherence risks producing what Couldry and Mejias (Reference Couldry and Mejias2019) call data colonialism: the subsumption of heterogeneity into extractable sameness. This dynamic is intensified by the fact that large language models and recommendation systems are trained disproportionately on dominant linguistic and cultural corpora. Marginalized voices may be underrepresented, averaged, or erased. The resulting synthetic discourse not only reproduces historical biases but also risks normalizing them as the baseline of reason itself.
Why Existing Frameworks Are Insufficient
The dynamics described above require a reconsideration of existing approaches to epistemic decay. Scholarship on cognitive bias, misinformation, platform governance, and ethical AI has generated important insights, but these frameworks often remain insufficient for describing a condition in which the infrastructures of communication themselves encode and automate cognitive distortion.
Approaches grounded in cognitive bias theory help to explain why certain cues, frames, and affects reliably structure judgment. Yet they often imply that rationality can be restored through improved information literacy, behavioral nudges, or individual debiasing (Van der Linden, Roozenbeek, and Compton Reference van der Linden, Roozenbeek and Compton2020). Such interventions include school-based media literacy curricula, accuracy prompts before sharing, fact checking, and debunking initiatives. The limitation is that they locate bias primarily in the individual mind rather than in the infrastructures that amplify, monetize, and automate bias at scale.
Similarly, models focused on filter bubbles and echo chambers identify important dynamics of cognitive homophily (Bakshy, Messing, and Adamic Reference Bakshy, Messing and Adamic2015; Pariser Reference Pariser2011; Sunstein Reference Sunstein2009). Yet exposure to diverse or excessive content can itself impair comprehension, memory, and reasoning under cognitive overload (Jeong and Hwang Reference Jeong and Hwang2016), while corrective interventions can trigger counterbias effects (Soprano et al. Reference Soprano, Roitero, La Barbera, Ceolin, Spina, Demartini and Mizzaro2024). The problem is therefore not only insufficient exposure to difference, but also the infrastructural organization of attention under conditions of overload.
Misinformation and disinformation studies tend to frame the problem as one of content accuracy or intentional deception, to be mitigated through fact checking, moderation, or labeling (Vosoughi, Roy, and Aral Reference Vosoughi, Roy and Aral2018; Wardle and Derakhshan Reference Wardle and Derakhshan2017). These approaches remain essential, but they often preserve a representational model of truth: the public sphere malfunctions because some statements are false. Opinionrot shifts the emphasis from false statements to decaying conditions of judgment. The problem is not only that publics encounter falsehoods, but also that the infrastructures through which they assess truth increasingly reward prediction, fluency, and engagement over verification.
Emerging work on technological debiasing offers partial responses (Reyero Lobo et al. Reference Lobo, Paula, Alani and Fernandez2023; Sundar, Liao, and Lee Reference Sundar, Liao, Lee, Bailey and Read2025) but faces structural limits. Cognitive biases are not simply pathologies; they are functional heuristics. Attempts to suppress them can produce fatigue, disengagement, or artificial forms of balance that distort democratic debate. Infrastructural debiasing, including ethical AI initiatives, seeks to reintroduce fairness, transparency, and accountability into information systems (Barocas, Hardt, and Narayanan Reference Barocas, Hardt and Narayanan2023). Yet bias audits and model interpretability often operate downstream of the profit motives driving engagement optimization. As critics of technological ethics argue, ethical discourse can itself function as a legitimating layer while leaving extractive infrastructures intact (Phan et al. Reference Phan, Goldenfein, Mann and Kuch2022).
These frameworks are therefore not wrong, but partial. They address symptoms of epistemic decay while often leaving intact the economic and technical architectures that produce it. Contemporary responses frequently treat the crisis of reasoning as a correctable deviation from an otherwise functional informational order (Danaher Reference Danaher2019). By contrast, enshittification, slopification, and automated murmuring suggest that the erosion of collective reasoning is not an episodic failure of cognition but a reconfiguration of information infrastructure itself.
Against this background, opinionrot should not be mistaken for individual cognitive failure or “bad thinking.” Rather, it represents a structural drift embedded in sociotechnical systems. The erosion of reasoning capacities emerges from the interplay of infrastructural incentives, automation, and information overload, not from irrational citizens. This aligns with theories of information decay that emphasize how feedback loops within complex media ecosystems tend toward efficiency rather than accuracy (Benkler Reference Benkler2006; DiResta Reference DiResta2018), and with concerns that new algorithmic and AI-generated information technologies may intensify existing information-consumption vulnerabilities (Shanmugasundaram and Tamilarasu Reference Shanmugasundaram and Tamilarasu2023).
Opinionrot is therefore best understood not as a replacement for existing frameworks, but as an integrative concept. It links cognitive capture, infrastructural decay, and discursive saturation within a broader account of how communicative systems deform the conditions of democratic judgment. It also avoids the overly bleak conclusion that public reason has disappeared altogether. More bounded institutional settings (such as some courts, expert bodies, investigative journalism, and slower editorial forums) can still sustain more traceable and revisable forms of judgment. These pockets of reasoning matter precisely because they reveal what opinionrot erodes: the institutional and communicative conditions under which opinion can remain contestable, accountable, and collectively revisable.
Why Opinionrot Intensifies Today
The preceding section identified three contemporary mechanisms through which opinionrot becomes infrastructural: platform decay, synthetic saturation, and automated public sense making. These mechanisms alter the conditions under which opinion is produced, circulated, and recognized as credible. Opinionrot intensifies today because older dynamics of persuasion, attention capture, and media concentration are now joined by generative systems capable of producing, ranking, and recirculating discourse at scale. The contemporary phase is therefore not merely a continuation of earlier media distortion. It marks a shift from the amplification of human expression to the partial automation of public sense making. Platforms no longer only mediate opinions that have been produced elsewhere; increasingly, they help to generate the communicative environment in which opinions appear plausible, visible, and socially validated. The following sections explain how this transformation can be diagnosed conceptually and illustrated empirically.
Mechanisms of Intensification
Opinionrot intensifies through three interrelated mechanisms: cognitive capture, infrastructural decay, and discursive saturation. Cognitive capture refers to the embedding of human biases (attentional, affective, and social) within digital infrastructures. Platforms exploit heuristics such as confirmation bias, availability bias, emotional salience, and negativity weighting because these facilitate rapid information selection in environments of abundance. Individuals use such heuristics to simplify complex informational and social environments, distinguishing allies from adversaries and assigning legitimacy to claims and sources. What begins as an adaptive cognitive shortcut increasingly becomes a computational feature. Engagement-based architectures amplify predictable reactions, transforming bias and social sorting from a limitation of reasoning into an economically valuable resource.
Infrastructural decay refers to the long-term deterioration of communicative systems under extraction-driven incentives. Doctorow’s (Reference Doctorow2025) account of enshittification and Zuboff’s (Reference Zuboff2019) analysis of surveillance capitalism both describe infrastructures that gradually subordinate public-facing functions to behavioral extraction and profit maximization. Under these conditions, visibility, credibility, and attention are increasingly organized around commercial imperatives rather than deliberative ones. Public reasoning becomes an externality rather than a purpose of communication systems.
Discursive saturation refers to the overload of meaning intensified by generative and high-volume communication environments. This mechanism has precedents in earlier mass-media and platform ecologies, but it acquires a distinctive form with synthetic content. As generative systems flood public discourse with statistically coherent but often semantically shallow material, the distinction between signal and noise becomes increasingly difficult to sustain. Slopification and automated murmuring generate what might be described as informational smog: a communicative environment in which abundance undermines rather than supports judgment.
These mechanisms do not have the same historical status. Cognitive capture and infrastructural decay have longer roots in persuasion, advertising, propaganda, media concentration, and platform capitalism. Discursive saturation, by contrast, is intensified by the recent scalability of generative media. The contemporary specificity of opinionrot lies in their combination: older systems for capturing attention and extracting behavioral data now operate alongside systems capable of producing synthetic discourse at scale.
Speed, Scale, and Automation
The intensification of opinionrot is inseparable from acceleration. Earlier communication systems certainly shaped public opinion, but they remained constrained by institutional bottlenecks, editorial processes, production costs, and temporal delays. Newspapers, broadcasters, and even early digital platforms operated within rhythms that still allowed some degree of verification, contestation, and revision.
Generative and predictive systems increasingly weaken those frictions. Synthetic content can be produced, personalized, and recirculated faster than institutions of verification, interpretation, and accountability can respond. Opinionrot is therefore not merely a problem of informational quantity but also of informational tempo. The speed of generation, circulation, and reaction increasingly exceed the speed of deliberation.
This acceleration transforms the relationship between judgment and communication. Under conditions of high-velocity circulation, public attention is repeatedly redirected before claims can be adequately assessed. The consequence is not necessarily misinformation, but a weakening of the social mechanisms through which beliefs are collectively revised. Publics become increasingly responsive while becoming less reflective.
From Amplified Opinion to Synthetic Plausibility
The contemporary phase of opinionrot also changes the relation between knowledge, verification, and credibility. Earlier media systems amplified, filtered, framed, or distorted human expression. Generative systems add something different: they produce discourse-like outputs that can appear socially grounded without being attached to accountable speakers, situated judgment, or traceable public contestation.
The transition from truth verification to pattern recognition marks a fundamental epistemic shift. Classical knowledge systems rested on the validation of sources, evidence, and expertise. By contrast, large-scale AI systems operate through the prediction of linguistic and behavioral continuities (Bender et al. Reference Bender, Braun, Möller and Gehrke2024; Floridi Reference Floridi, Miller and Taddeo2020). Large language models function as probability engines: they generate outputs according to statistical coherence rather than factual verification. Their authority often derives from fluency rather than evidence (Birhane Reference Birhane2021).
As Kockelman (Reference Kockelman2024) argues, such systems do not simply communicate discourse but also generate discourse-like outputs from prior linguistic regularities. Weatherby and Justie (Reference Weatherby and Justie2022) similarly show how these systems can simulate socially grounded language without participating in the accountable and situated practices through which public judgment ordinarily becomes meaningful.
The significance of this transformation is not simply that AI may produce falsehoods. Rather, it alters the conditions through which credibility itself is established. In McLuhan’s ([1964] Reference McLuhan1994) terms, the medium once again becomes the message. The process of generation increasingly substitutes for the process of justification. Public discourse risks shifting from agonistic pluralism toward forms of algorithmic consensus in which disagreement is smoothed into statistical averages. If democratic reasoning depends upon exposure to diverse standpoints (Mouffe Reference Mouffe2013), predictive systems may privilege coherence over contestation.
This concern is reinforced by Couldry and Mejias’s (Reference Couldry and Mejias2019) account of data colonialism. Large language models and recommender systems are trained disproportionately on dominant linguistic and cultural corpora. Marginalized voices are frequently underrepresented, averaged, or rendered statistically insignificant. Synthetic discourse can therefore reproduce historical inequalities while presenting them as neutral outputs of computational reason.
Diagnosing Opinionrot
If opinionrot names a structural erosion of collective reasoning, its detection cannot rely exclusively on conventional measures of misinformation, belief accuracy, or individual cognition. These indicators capture important symptoms but not the underlying transformation of communicative infrastructures. Opinionrot becomes visible through changes in three domains: infrastructure, discourse, and social cognition.
At the infrastructural level, indicators include concentration of visibility, declining source diversity, linguistic homogenization, and growing dependency upon a small number of platforms for information discovery. Such measures reveal the narrowing of communicative pathways and the concentration of epistemic power.
At the discursive level, indicators include rising proportions of synthetic content, increasing engagement–veracity asymmetries, and declining signal-to-noise ratios. Slopification manifests not only as misinformation but also as repetition, semantic shallowness, and the proliferation of content detached from identifiable authorship and accountability.
At the sociocognitive level, indicators include narrowing attention diversity, growing news fatigue, declining trust, the increasing volatility of belief formation, and shortening cycles of reaction. These measures do not capture individual irrationality but rather the weakening of collective capacities for sustained judgment.
No single metric can capture opinionrot. Rather, the concept points toward a composite diagnosis that links infrastructural organization, discursive environments, and cognitive outcomes. The objective is not to identify false beliefs but to assess the conditions under which beliefs can be publicly examined, challenged, and revised. The following comparative illustration demonstrates how such a diagnostic framework might be applied to a single political controversy across different media environments.
A Comparative Illustration: Equality and Affirmative Action
The long-running debate over equality and affirmative action illustrates how the same political controversy can unfold under different communicative conditions. The example is useful not because earlier debates were free of bias, distortion, or inequality, but because it allows comparison across media environments. It helps to distinguish persistent disagreement over fairness, merit, quotas, representation, and historical disadvantage from the structural conditions captured by opinionrot.
A cross-temporal study could examine how debates over fairness, merit, quotas, and equality evolved across newspaper editorials, televised debates, policy documents, social media platforms, and AI-mediated information environments between 1970 and 2025. Key moments might include United States Supreme Court rulings on affirmative action, French parité reforms, and European diversity directives. Because the underlying controversy persists across these periods, the comparison isolates how changing media infrastructures alter the form, tempo, visibility, and revisability of public disagreement. The purpose would be to trace a shift from controversies mediated primarily through identifiable institutions and speakers toward controversies increasingly organized through algorithmic ranking, engagement signals, synthetic summaries, and predictive systems.
Such a study could operationalize the indicators discussed above. Infrastructural analysis would examine the concentration of visibility and platform dependency. It could also assess source diversity, recommender dependency, and the extent to which a small number of platforms determine which interpretations of affirmative action become publicly salient. Discursive analysis could identify lexical homogenization, engagement–veracity asymmetries, and the repetition of emotionally charged frames such as “reverse discrimination,” “meritocracy,” or “quotas.” It could further examine whether these frames circulate as parts of attributable and extended arguments or as decontextualized fragments optimized for rapid recognition and reaction. Sociocognitive indicators could examine trust, deliberative latency, and the speed with which controversies become polarized. Additional indicators might include attention diversity, news fatigue, belief volatility, and the degree to which users infer public consensus from likes, rankings, recommendations, generated summaries, and other algorithmically organized signals.
The point is not that debates in the 1980s or 1990s were normatively ideal; they were often exclusionary, unequal, and highly contentious. Nor should slower or more institutionalized communication be treated automatically as more democratic. Established media and political institutions can marginalize minority perspectives and protect dominant understandings of equality. Yet many of these earlier debates were more traceable, slower moving, and more institutionally mediated than contemporary platform discourse. Arguments were more frequently linked to identifiable authors, editorial processes, policy documents, and judicial reasoning. Contestation remained imperfect, but revision was often easier. The object of disagreement was also frequently more stable: participants could respond to a particular ruling, editorial, speech, dataset, or legislative proposal rather than to a continuously shifting mixture of clips, rankings, reactions, and synthetic interpretations.
In contemporary platform and AI-mediated discourse, the tempo of circulation can exceed the tempo of judgment. Users may encounter a decontextualized quotation, a trending interpretation, or a generated summary before they encounter the source (if any) from which it derives. Public attention may then move to a new controversy before evidentiary disputes, conceptual disagreements, or normative claims have been clarified. The result is not necessarily more falsehood, but a communicative environment in which reaction becomes easier while sustained revision becomes harder. Synthetic saturation intensifies this problem because credible journalism, primary sources, and reasoned argument must compete with a potentially unlimited supply of fluent, cheap, and endlessly reproducible approximations.
This comparison clarifies the normative standard that underlies the concept of opinionrot. The concern is not the disappearance of consensus or disagreement. Misinformation analysis might ask whether a specific assertion about affirmative action is accurate; polarization research might ask whether opposing groups are moving further apart; cognitive bias research might ask why individuals prefer some arguments. Opinionrot asks how the communicative environment organizes visibility, plausibility, attention, and social inference in ways that weaken the shared capacity to justify, challenge, and revise judgments. It is the erosion of communicative conditions that allows disagreements to remain traceable, contestable, and revisable over time. Opinionrot therefore describes not the end of public debate but the gradual weakening of the infrastructures that make public reasoning possible. In the affirmative action example, this weakening would become visible not because disagreement persists, but because the surrounding infrastructure increasingly manufactures salience, rewards affective repetition, obscures provenance, simulates public sentiment, and accelerates circulation beyond the temporal capacities of verification and deliberation.
This remains a conceptual illustration rather than empirical proof. Demonstrating opinionrot would require systematic cross-temporal evidence and attention to countervailing developments, including expanded participation, greater access to marginalized perspectives, and institutional settings in which claims remain accountable and revisable. The utility of the example is therefore diagnostic: it shows how the concept connects changes in political economy, media infrastructure, discourse, and cognition within an example controversy.
Theoretical Implications
Rather than treating epistemic crisis as a moral failure, a decline in individual rationality, or the collapse of an otherwise stable informational order, the concept directs attention to the infrastructures through which public judgment is organized. The argument is not that truth has disappeared, but that the conditions through which truth claims acquire visibility, credibility, and authority have been transformed.
This perspective differs from accounts that focus primarily on misinformation, polarization, or cognitive bias. Such phenomena remain important, but they often appear as symptoms of a broader structural transformation. Opinionrot shifts the level of analysis from individual beliefs and discrete content to the sociotechnical environments within which reasoning takes place. The central claim is that collective reasoning can deteriorate even when information remains abundant, because the infrastructures responsible for organizing attention, visibility, and interpretation increasingly prioritize prediction, engagement, and behavioral extraction over deliberative functions.
The concept therefore contributes to ongoing debates about democratic communication in three ways. First, it offers an integrative framework. Existing approaches frequently focus on isolated dimensions of epistemic decline: cognitive bias, misinformation, platform governance, AI ethics, propaganda, attention economies, or democratic polarization. Opinionrot brings these literatures into the conversation by locating them within a broader account of how communicative infrastructures progressively absorb and operationalize cognitive vulnerabilities.
Second, it shifts attention from content to conditions. Public debate is often framed in terms of whether particular claims are true or false. Opinionrot instead asks under what conditions claims can be publicly examined, contested, and revised. This shift places emphasis on the environments that make reasoning possible rather than solely on the outputs of those environments.
Third, it provides a vocabulary for analyzing the transition from mediated publics to predictive publics. Earlier communication systems primarily selected, filtered, and amplified human expression. Contemporary generative systems increasingly participate in the production of discourse itself. The result is not simply more communication, but a reorganization of how plausibility, authority, and collective judgment are formed.
At the same time, the concept should not be interpreted as a claim that democratic reasoning has disappeared. Such a conclusion would reproduce the deterministic logic that this article seeks to avoid. Public reasoning persists within many institutional settings, including investigative journalism, courts, scientific communities, professional associations, civil society organizations, and other spaces where claims remain traceable, contestable, and revisable. These settings demonstrate that opinionrot is neither universal nor total. Rather, it describes the growing structural pressures that make such forms of reasoning more difficult to sustain. The argument does not seek a return to an imagined age of consensus, objectivity, or epistemic purity. Democratic life has always involved conflict, persuasion, inequality, and competing interpretations of reality. The concern is instead with preserving the communicative conditions that allow disagreement to remain productive rather than merely reactive. Public reasoning depends not on consensus, but on the continued possibility of justification, contestation, and revision.
From this perspective, the challenge posed by opinionrot is ultimately ecological. Just as environmental degradation results from the overexploitation of natural resources, epistemic degradation emerges from the overexploitation of attention, cognition, and communicative capacity. The task is not simply to correct false information, but to sustain the informational environments within which collective reasoning can continue to operate.
Epilogue
This essay began with a paradox: communication technologies designed to expand access to information can also weaken the conditions that make collective reasoning possible. To understand this paradox, the article traced a conceptual-historical trajectory from persuasion, to attention capture, to prediction. Across these transformations, the central argument has been that opinionrot emerges when cognitive vulnerabilities become progressively embedded within the infrastructures that organize public communication.
The concept of opinionrot does not describe a sudden rupture. Nor does it imply that contemporary societies have entered an unprecedented age of irrationality. Public life has always been shaped by persuasion, unequal power, strategic communication, and informational asymmetries. What distinguishes the present moment is the increasing capacity of digital and generative systems to automate, scale, and normalize these dynamics within the ordinary operation of communication infrastructures.
The argument has also sought to avoid a second temptation: nostalgia. The purpose of the concept is not to recover a lost golden age of public reason. Earlier communication systems were themselves marked by exclusion, hierarchy, propaganda, and unequal access to voice. The relevant comparison is therefore not between a rational past and an irrational present, but between different configurations of communicative infrastructure and their consequences for collective judgment. Opinionrot arises when the communicative conditions that make those disagreements publicly traceable, contestable, and revisable are progressively weakened.
Seen in this light, opinionrot is best understood as a diagnostic concept. It identifies a structural tendency rather than a completed condition. It names the processes through which reasoning becomes increasingly difficult to sustain when communication systems optimize for prediction, engagement, and extraction. Its analytical value lies in directing attention to the infrastructures that organize visibility, plausibility, and public judgment. Yet tendencies are not destinies. The persistence of investigative journalism, scientific inquiry, legal reasoning, public interest institutions, and democratic contestation demonstrates that collective reasoning remains possible, even under increasingly unfavorable conditions.
The central question raised by opinionrot is therefore not whether truth survives automation: it is whether societies can preserve and cultivate the institutional, technological, and communicative arrangements through which claims remain open to scrutiny, disagreement, and revision. The future of democratic reasoning depends less on eliminating uncertainty than on sustaining those conditions.
If the concept of opinionrot has value, it is because it provides a language for recognizing the pressures currently acting upon the conditions of collective reasoning. It is also useful for understanding why the defense of collective reasoning increasingly requires attention not only to what information says, but also to the infrastructures through which it becomes thinkable.
Acknowledgments
I thank Professor Davide Morselli (University of Lausanne, Switzerland) for insightful discussions that helped to refine and finalize the conceptual framework. No funding is associated with this research, and there are no conflicts of interest to report.
Use of Artificial Intelligence (AI)
AI-assisted writing tools, including Grammarly, were used to support the drafting and editing of portions of this article. All generated or suggested text was carefully reviewed, revised where necessary, and verified by the author for accuracy, originality, and compliance with the journal’s standards.
Open access
