How Algorithms Turn Outrage Into Revenue: The Rage Bait Economy

A multidisciplinary analysis of rage bait: its psychological mechanisms, its algorithmic feedback loop, and its implications for the governance of social platforms

Table of Contents

Introduction

Section 1. Psychological and social foundations of rage bait

• 1.1. The negativity asymmetry in digital environments: why anger spreads faster and more intensely than positive emotion
• 1.2. Moral outrage as a signal of group belonging and identity positioning
• 1.3. The role of social validation and emotional contagion in amplifying provocative content
• 1.4. The paradox of critical engagement: sharing to debunk as a vector of virality

Section 2. Algorithmic architecture and the attention economy

• 2.1. Design principles of recommender systems: maximizing time-on-platform and prioritizing intense interactions
• 2.2. Empirical evidence: engagement-based algorithms systematically amplify moral outrage relative to chronological feeds
• 2.3. Rage bait as a rational monetization strategy for content creators: economic incentives and performance-based pay
• 2.4. AI-generated rage bait: a new stage in the dehumanization of provocation

Section 3. Systemic consequences and social costs

• 3.1. Political polarization and the fragmentation of the public sphere: outrage as a catalyst for filter bubbles and echo chambers
• 3.2. Misinformation and the erosion of epistemic trust: how rage bait spreads false news by disabling verification
• 3.3. Impact on mental health and digital well-being: doomscrolling, emotional fatigue, and impaired affect regulation
• 3.4. The "noisy minority" problem: the concentration of influence among extreme users and its distortion of perceived social norms

Section 4. Institutional, regulatory, and design responses

• 4.1. Algorithmic redesign proposals: interventions on the perception of social norms and de-escalating outrage
• 4.2. The limits of self-regulation: why technical tweaks are not enough to defuse structural toxicity
• 4.3. The emerging regulatory framework: the EU's Digital Services Act (DSA) and its requirements for algorithmic transparency and systemic risk management
• 4.4. Looking ahead: toward an attention economy built on well-being rather than extreme emotional intensity

Bibliography

Introduction

In December 2025, Oxford University Press announced the selection of rage bait as its word of the year, defining it as "content that is deliberately designed to provoke anger or outrage by being annoying, provocative, or offensive, typically for the purpose of increasing traffic or engagement on a webpage or piece of social media content" (8). The decision, backed by more than 30,000 public votes, reflects not only the term's penetration into everyday language, but a growing awareness of a phenomenon that has reshaped the digital attention economy (19).

Rage bait is an evolution of classic clickbait. Where clickbait appeals to curiosity to drive clicks, rage bait exploits outrage as an engagement driver. Its effectiveness rests on a combination of psychological vulnerabilities, economic incentives, and algorithmic bias that have turned anger into one of the most reliable currencies in the digital ecosystem (1, 3). The phenomenon is not marginal: creators who earn most of their income from direct platform payouts turn to rage bait as a core monetization strategy (13), while recommendation systems systematically amplify morally outrageous content over neutral or positive content (5).

This article offers a multidisciplinary analysis of rage bait, drawing on social psychology, algorithmic economics, digital communication, and platform governance. Its aims are twofold: first, to unpack the mechanisms that make rage bait so effective from the creator's standpoint and so compelling from the user's; second, to assess the institutional and regulatory responses currently being tested to mitigate its most harmful effects, with particular attention to the European Union's Digital Services Act (DSA) and to algorithmic redesign experiments aimed at a less polarized attention economy.

Section 1. Psychological and social foundations of rage bait

1.1. The negativity asymmetry in digital environments: why anger spreads faster and more intensely than positive emotion

The effectiveness of rage bait rests on a deep psychological substrate: negative emotions, and anger in particular, spread significantly faster and more intensely than positive emotions in digital environments (7). This phenomenon, documented in the literature on emotional contagion on social media, answers to an evolutionary imperative: from an adaptive standpoint, it is more urgent for an individual and their group to process and spread information about a threat or an injustice than about a pleasurable experience (7). Human attention is biased toward what can harm us, and platform algorithms have learned to exploit this ancestral bias.

Research in cognitive psychology has shown that emotionally charged content — especially content that provokes outrage — triggers physiological and neurological responses that increase the likelihood of immediate interaction (18). When a user encounters a post that angers them, the limbic system responds before the prefrontal cortex can exert inhibitory control. This cognitive short-circuit is exactly what rage bait creators deliberately exploit: the gap between perceiving the offense and reacting (commenting, sharing, reacting in anger) is short enough to bypass verification and critical reflection. Source (17) systematizes this anatomy of the phenomenon, identifying a recurring pattern in the proliferation of rage bait: the provocation is built on deliberately simplified or exaggerated claims, whose function is not to convey accurate information but to maximize the probability of an immediate visceral reaction.

The literature likewise notes that content provoking moral outrage spreads at least as strongly as trustworthy news, and sometimes faster, because people share outrage-inducing misinformation without verifying its accuracy (6). This behavior is not simple cognitive negligence but a social function: sharing content that provokes anger is a way of signaling group belonging, of stating a moral position, and of reinforcing community bonds against a perceived threat (6). Sharing thus becomes a marker of identity, while the truthfulness of the content becomes secondary.

1.2. Moral outrage as a signal of group belonging and identity positioning

Rage bait does not operate in a vacuum; its effectiveness depends on the existence of interpretive communities that share values, norms, and moral sensibilities. Moral outrage acts as social glue: when an individual shares or reacts angrily to a post, they are signaling to their network which group they belong to and where their ethical boundaries lie (6). This mechanism, studied within social psychology, explains why rage bait tends to polarize rather than unify: by appealing to group identity, it reinforces the boundary between "us" and "them."

Source (18) adds that content directed at opposing social groups, or shared by influential users, amplifies its own virality, polarizing communities and intensifying moral outrage. This effect is reinforced by the architecture of the platforms themselves, which allow audiences to be segmented and emotionally charged messages to be targeted at specific niches. Rage bait is therefore not a universal message: it is a message designed to resonate with the values and grievances of a particular audience, which multiplies its capacity to provoke a visceral response.

1.3. The role of social validation and emotional contagion in amplifying provocative content

Social networks are, above all, ecosystems of validation. Every interaction — a "like," a comment, a repost — acts as a reinforcement signaling to both the creator and the algorithm that the content is relevant. Emotional contagion, understood as the transfer of affective states between individuals through the observation of emotional expressions, is magnified in these environments by repeated exposure to posts that generate anger (18). The more an outrageous piece of content is shared, the more users see it, and the more users react with anger — creating an emotional feedback loop that algorithms not only register but actively amplify.

Source (1) introduces a key concept: the "manufacture of dissent" and "moral outrage networks." Rage bait does not merely reflect pre-existing outrage; it builds it, shapes it, and intensifies it through framing choices, the omission of nuance, and the exaggeration of conflict. In this sense, rage bait is a technology for the production of emotion, not a mere mirror of reality. And this production is facilitated by platform architecture, which offers real-time metrics (views, comments, reactions) that feed the impression that outrage is widely shared and dominant, even when it is not (16).

1.4. The paradox of critical engagement: sharing to debunk as a vector of virality

One of the more counterintuitive findings in the study of rage bait is that critical engagement — that is, sharing or commenting on content with the intent of debunking or denouncing it — contributes just as much to its virality (1). Algorithms do not distinguish between positive and negative intent; they only measure activity. A user who shares a piece of rage bait along with a comment refuting it generates the same traffic and the same engagement as one who shares it in agreement. Source (1) calls this "the debunking paradox": the attempt to fight misinformation or provocation through direct engagement ends up strengthening it, because the algorithmic signal sent is identical.

This has profound implications for moderation and digital literacy efforts. Fact-checking campaigns circulated on the platforms themselves, if not carefully designed, can become unwitting vehicles of amplification. Source (6) confirms that sharing outrage-inducing misinformation without verifying its accuracy is a common practice, precisely because the act of sharing carries a social and emotional value that precedes cognitive evaluation. Breaking this loop therefore requires interventions that target not only content, but the structural incentives that reward emotional intensity over accuracy.

Section 2. Algorithmic architecture and the attention economy

2.1. Design principles of recommender systems: maximizing time-on-platform and prioritizing intense interactions

The recommender systems that govern the major social platforms — TikTok, X, Facebook, Instagram — are not neutral with respect to the content they distribute. Their architecture answers to a primary objective: maximizing time-on-platform and, with it, advertising revenue (3). To achieve this, algorithms prioritize posts that generate fast, intense interactions, since these are the strongest predictors of user retention. Within this framework, emotionally charged content — and especially content that provokes anger or outrage — enjoys a structural competitive advantage over neutral or positive content (1).

Source (1) develops the concept of an "outrage economy" to explain how platforms have internalized this bias. Algorithm engineers do not explicitly design systems to promote anger; what they design are systems that optimize engagement, and engagement turns out to be robustly correlated with negative emotional intensity. This correlation is not accidental: negative emotions, anger in particular, trigger physiological responses that increase the likelihood of immediate interaction (commenting, resharing, reacting) and lower the threshold for impulsive action (3). As a result, rage bait becomes an algorithmically favored product without any explicit intent behind it. It is, rather, an emergent effect of optimizing shallow metrics. Source (14) reaches a convergent conclusion from a more historical angle: every technological leap in media — from print to the telegraph, from radio to television — has produced a period of disruption in society's capacity to process information, and today's social networks are going through their own "dark valley," one in which the architecture of economic incentives systematically favors content that polarizes over content that informs.

Source (3), focused on the digital sports ecosystem, provides empirical evidence of how rage bait strategies are directly rewarded by algorithms. In its study, posts appealing to rivalry, refereeing controversy, or confrontation between fanbases generated significantly higher engagement rates than those offering technical analysis or neutral reporting. Creators adopting these strategies saw their visibility — and therefore their earnings — increase, creating a perverse incentive to escalate provocation.

2.2. Empirical evidence: engagement-based algorithms systematically amplify moral outrage relative to chronological feeds

Quantitative evidence of algorithmic bias toward outrage has grown considerably in recent years. A recent study (5), conducted during a national election campaign, compared the performance of engagement-based algorithmic feeds against strictly chronological feeds. The results were conclusive: algorithmic feeds systematically amplified toxic and morally outrageous content relative to chronological feeds, with the largest increases recorded in the categories of moral outrage and polarizing political content (5).

This finding is crucial because it dismantles the narrative that algorithms simply reflect user preferences. If users wanted to see outrageous content, a chronological feed would show it too, but in proportion to its posting frequency. The fact that algorithmic feeds disproportionately amplify that content indicates that recommender systems are not passive mirrors but active amplifiers of emotional intensity (5). The study's authors suggest that this amplification occurs because algorithms learn to detect weak engagement signals (such as dwell time or scroll speed) and to prioritize content that maximizes those signals, even when users themselves report preferring less polarizing content.

Source (16) adds a further layer of complexity: algorithms designed to maximize engagement end up entrenching the dominance of the loudest, most extreme users. A small number of high-profile users capture most of the attention and influence, and these users tend to produce more polarized, provocative content precisely because it performs best within the algorithmic ecosystem. The result is a systematic distortion of perceived social norms: users come to believe that outrage and conflict are more common than they actually are, which in turn fuels further outrage (16).

2.3. Rage bait as a rational monetization strategy for content creators: economic incentives and performance-based pay

For content creators, rage bait is neither a pathology nor an accident; it is a rational revenue-maximization strategy. Source (13), from Stanford Report, notes that rage bait is particularly prevalent among creators who earn most of their income from direct platform payouts, such as TikTok's creator funds or YouTube's ad-revenue-sharing systems. These systems reward view count, watch time, and engagement rate — metrics that rage bait generates with unusual efficiency.

The economic logic is simple: producing content that generates anger costs the same as producing neutral or positive content, but its returns in terms of reach and revenue are disproportionately higher (13). Moreover, rage bait tends to be cheaper to produce than high-quality informational or artistic content, since it requires no research, fact-checking, elaborate production, or specialized creative talent. A provocative frame, a controversial claim, or a confrontational narrative is enough to activate the outrage circuit.

Source (1) delves further into the political economy of rage bait, describing how the "manufacture of dissent" has become a business model. Creators do not merely respond to existing demand for outrageous content; they actively stimulate it through framing strategies, selective presentation of information, and appeals to group identity. Rage bait is, in this sense, a carefully engineered product designed to exploit the emotional vulnerabilities of audiences, and its production responds to clear, measurable economic incentives.

2.4. AI-generated rage bait: a new stage in the dehumanization of provocation

A recent and particularly worrying evolution of the phenomenon is the automated generation of rage bait through artificial intelligence systems. Source (2), published in New Media & Society (a SAGE journal), presents a case analysis of AI-generated rage bait, identifying it as a severe form of automated trolling. Unlike human-created rage bait, AI-generated rage bait has no human actor behind it who can be held accountable, nor does it respond to a genuine communicative intent — it responds to the optimization of language models trained to maximize engagement metrics (2).

The study (2) documents how large language models can generate provocative content at industrial scale, tailored to specific audiences and cultural contexts. These systems are able to identify successful patterns in prior posts and replicate them with enough variation to evade duplicate-content detection systems. Automating rage bait drastically reduces production costs and enables a scalability that human creators cannot match.

Source (2) warns that this automation introduces new challenges for content moderation and platform governance. If human-generated rage bait was already difficult to address, machine-generated rage bait — which can be produced in massive volumes and adapt dynamically to algorithmic changes — poses a qualitatively different threat. The dehumanization of provocation removes even the ethical limit that some human creators might otherwise impose on themselves, turning outrage into an industrial byproduct of data processing (2).

Section 3. Systemic consequences and social costs

3.1. Political polarization and the fragmentation of the public sphere: outrage as a catalyst for filter bubbles and echo chambers

One of the best-documented consequences of rage bait and its algorithmic amplification is the intensification of political polarization and the fragmentation of the digital public sphere. Source (18) notes that content shared by influential users, or directed at opposing social groups, amplifies its own virality, polarizing communities and intensifying moral outrage. This process is not incidental: rage bait deliberately appeals to group identities, reinforcing the boundary between "us" and "them" and turning public debate into a confrontation between emotional tribes.

The underlying mechanism is that of filter bubbles and echo chambers. By prioritizing content that generates interaction, algorithms tend to expose users to more content that confirms their biases and stokes their outrage, while filtering out nuanced or dissonant perspectives. Source (5) provides robust empirical evidence: during an election campaign, algorithmic feeds systematically amplified moral outrage and polarizing political content relative to chronological feeds. This means that algorithmic architecture does not merely reflect existing polarization — it accelerates and deepens it by rewarding the most extreme, emotionally intense expressions.

Source (16) adds that this process distorts the perception of social norms. Users repeatedly exposed to outrageous, polarized content come to believe that conflict and hostility are more common and more intense in the general population than they actually are. This mistaken perception, in turn, legitimizes and encourages more extreme behavior, creating a feedback loop that erodes the foundations of democratic debate: the willingness to listen, respect for dissent, and the search for common ground.

3.2. Misinformation and the erosion of epistemic trust: how rage bait spreads false news by disabling verification

The relationship between rage bait and misinformation is close and concerning. Source (6) shows that moral outrage helps spread misinformation at least as strongly as trustworthy news. The central finding is that people share outrage-inducing misinformation without verifying its accuracy because the act of sharing serves a social function of signaling group belonging and moral positioning (6). Fact-checking becomes a cognitive cost that many users are willing to skip when the emotional reward of sharing is high enough.

This has profound implications for epistemic trust — that is, trust in the institutions and processes that generate reliable knowledge. When misinforming rage bait circulates at the same speed as, or faster than, verified information, the distinction between true and false erodes, and a generalized skepticism sets in that affects traditional media as well as scientific and governmental sources alike. Source (1) frames this as part of the "manufacture of dissent": rage bait does not merely spread false information — it creates an environment in which any information is suspect and any claim can be challenged by appealing to emotion rather than evidence.

The automation of rage bait through AI, documented in (2), aggravates this problem by scaling the production of emotionally charged misinformation. Generative models can produce variations of false content tailored to different audiences and contexts, making verification and moderation efforts even harder.

3.3. Impact on mental health and digital well-being: doomscrolling, emotional fatigue, and impaired affect regulation

The psychological costs of sustained rage bait consumption are beginning to draw clinical and scientific attention. Source (12) examines how rage bait, doomscrolling (compulsively scrolling through negative news), and algorithmic amplification quietly alter attention, emotional regulation, and user perception. Constant access to live reactions and content designed to provoke anger creates an atmosphere of perpetual intensity that makes disconnection and healthy emotional processing difficult (12).

Doomscrolling is a closely related phenomenon: users, caught in loops of outrage, consume large volumes of negative content with no clear informational purpose, simply because the algorithm keeps serving it to them. This behavior is associated with symptoms of anxiety, irritability, cognitive fatigue, and a generalized sense of hopelessness. Source (12) suggests that rage bait acts as a chronic trigger of stress responses, repeatedly and prolongedly activating the sympathetic nervous system, with cumulative effects on physical and mental health.

In addition, impaired affect regulation manifests as reduced frustration tolerance and heightened emotional reactivity in offline contexts (12). Users accustomed to the intensity of rage bait may find everyday exchanges bland or irrelevant, or may react with disproportionate hostility to minor disagreements. This spillover from digital dynamics into real life is one of the less visible but more insidious consequences of the phenomenon.

3.4. The "noisy minority" problem: the concentration of influence among extreme users and its distortion of perceived social norms

Source (16) introduces a critical finding for understanding the systemic dynamics of rage bait: a small number of high-profile users capture most of the attention and influence on platforms. These users, who tend to produce more polarized and provocative content, dominate the digital public sphere not because they are representative of the population, but because algorithms designed to maximize engagement disproportionately amplify their posts (16).

The result is a systematic distortion of the perception of social norms. Source (16) calls this the "noisy minority problem": the most extreme and combative users capture attention, creating the illusion that their views are more common and more intense than they really are. Moderate users, seeing the space dominated by radical content, may withdraw from public conversation altogether or become radicalized themselves in order to be heard — pushing the entire discursive spectrum toward the extremes.

This phenomenon is compounded by AI-generated rage bait (2), which can artificially amplify the noisy minority through automated accounts that simulate support or outrage, further distorting the perception of public opinion. The combination of algorithms that favor intensity, creators who exploit that favoritism, and bots that scale the process turns the digital public sphere into a stage where representativeness gives way to emotional spectacle.

Section 4. Institutional, regulatory, and design responses

4.1. Algorithmic redesign proposals: interventions on the perception of social norms and de-escalating outrage

The accumulated evidence on the distorting effects of engagement-based algorithms has spurred growing interest in redesign proposals that mitigate these negative effects without sacrificing the user experience. A recent milestone in this direction is the study published in Nature (2026) by Brady and colleagues, the first large-scale experiment with full control over recommendation algorithms in a real-world environment (5). The researchers built custom feed-ranking algorithms and randomly assigned 2,000 participants to use them for eight weeks, spanning the period before and after the 2024 US presidential election (5).

The study's results are conclusive. Engagement-based feeds systematically amplified intergroup, moralized, and emotional (IME) content and toxic content relative to reverse-chronological feeds, with the largest increases recorded in moral outrage and political content (5). Specifically, the amplification of moral outrage and political content rose by roughly 37% before the election and nearly 80% after it, relative to the chronological-feed baseline (5). Engagement-based feeds also reduced the accuracy of social-norm perception and increased perceived partisan animosity (5).

In response to this diagnosis, the researchers designed and tested a "diversified extremity algorithm," conceived to reduce the disproportionate influence of extreme users (5). This algorithm operated through three mechanisms: downranking users who posted with excessive frequency ("super-posters"), lowering the probability that toxic posts would appear, and increasing the probability of constructive dialogue (5). The results were significant: the diversified extremity algorithm reduced exposure to IME and toxic content, improved the accuracy of social-norm perception, and — crucially — maintained comparable levels of platform enjoyment (5). This finding suggests that reducing the influence of extreme users can correct algorithmic distortions without degrading the user experience (5). As lead researcher William Brady put it, "for years, the debate over algorithmic amplification has been hampered by the fact that independent researchers don't have full control over the algorithms they study. Bluesky's open architecture let us build the algorithms ourselves, run a real-world experiment at scale, and measure what these systems actually do to political conversation" (5).

4.2. The limits of self-regulation: why technical tweaks are not enough to defuse structural toxicity

Despite these advances in algorithmic redesign, there is growing consensus that purely technical solutions, voluntarily implemented by the platforms themselves, are insufficient to address the structural toxicity of the digital ecosystem. Sources (9, 20) argue that no adjustments can make social media less toxic, given that social media, in its fundamental architecture, fosters outrage, enables radicalization, and generates filter bubbles. This radical position finds support in source (16), which argues that social media likely cannot be "fixed" on its current terms: a small number of high-profile users capture most of the attention and influence, and algorithms designed to maximize engagement end up amplifying outrage and conflict, entrenching the dominance of the loudest, most extreme users.

The problem of self-regulation is, at bottom, a problem of incentives. Platforms are for-profit companies whose business model depends on maximizing time-on-platform and engagement. As source (3) documents, rage bait strategies are directly rewarded by algorithms, increasing revenue for both creators and platforms. Asking these companies to voluntarily reduce the amplification of content that generates profit for them amounts to asking them to act against their own economic interest. Source (1) describes this phenomenon as the "outrage economy": anger has become one of the most reliable currencies of the digital economy, and recommender systems have learned to prioritize content that triggers the strongest emotions (3).

Source (15) adds that attention has become a "geoeconomic asset": platforms optimize for engagement and time spent, while political and media actors compete in the same market. In this context, voluntary self-regulation collides with the competitive logic of the market: a platform that reduces the amplification of outrageous content risks losing users and revenue to competitors that do not. This prisoner's dilemma makes external regulatory intervention not merely convenient, but necessary.

4.3. The emerging regulatory framework: the EU's Digital Services Act (DSA) and its requirements for algorithmic transparency and systemic risk management

The most ambitious regulatory response to the algorithmic amplification of outrage is the European Union's Digital Services Act (DSA), a regulation adopted in October 2022 that sets uniform rules for digital services (0, 10). The DSA represents a paradigm shift in platform governance: its goal is to create a safer online environment where fundamental rights, including freedom of expression and information, are protected (10). The DSA's fundamental novelty is that its focus is not limited to individual illegal content, but extends to so-called "systemic risks" tied to the very design of services, including recommender systems (2, 11).

Under the DSA's logic, recommender systems are responsible for suggesting, ranking, and prioritizing information, and have a significant impact on users' ability to retrieve and interact with information online (10). Providers of very large online platforms (VLOPs) and very large online search engines (VLOSEs) — those exceeding 45 million monthly active users in the EU — are required to assess and mitigate identified systemic risks under Articles 34 and 35 of the DSA (0, 10). These risks include, among others, the spread of illegal content, the impact of algorithms on public debate, electoral processes, and minors' mental health (0, 2). Recommender systems must receive particular attention in this assessment, since they play a significant role in amplifying certain content, in the viral spread of information, and in stimulating online behavior (10).

The DSA also introduces concrete instruments to make this oversight effective. Article 40 establishes, for the first time, a legal right for independent researchers to access platform data in the public interest (11). Once designated as VLOPs or VLOSEs, platforms must provide data access to support research on "systemic risks" (11). Article 40 creates two pathways: Article 40(12) allows access to publicly available data beyond platforms' voluntary tools, while Article 40(4) allows vetted researchers to request non-public data — such as exposure logs, moderation records, and recommendation metrics — through national Digital Services Coordinators rather than from the platforms themselves (11). Both pathways are limited to the study of systemic risks and, in the case of Article 40(4), of mitigation measures (11).

The European Commission has backed this framework with technical initiatives such as the creation of the European Centre for Algorithmic Transparency (ECAT), launched to support DSA enforcement with technical expertise and to conduct research on the long-term impact of algorithmic systems (10). The Commission's Joint Research Centre (JRC) has likewise proposed a taxonomy of study designs for auditing algorithmic risks, including categories such as risk-discovery studies, reverse-engineering studies, interface-design studies, and risk-measurement studies (10). These tools are intended to give regulators and the research community the means to rigorously assess the impact of algorithms on society.

The European approach is not the only one available, and the contrast with other jurisdictions illuminates its distinctive features. The United Kingdom, following its departure from the Union, has opted for a structurally different model under the Online Safety Act 2023: rather than the DSA's quantitative threshold of 45 million users triggering an enhanced regime, the UK regulator, Ofcom, classifies services by functional risk category and requires annual risk assessments that, since 2025, must expressly address the design of recommender systems and personalized feeds (21). The first assessment cycle, covering 2025, revealed significant gaps in the information providers themselves supplied about algorithmic governance, prompting Ofcom to strengthen its documentary requirements for the second cycle (21). The comparison is doctrinally instructive: while the European DSA structures algorithmic oversight around platform size and a right of access for external researchers, the British model relies on a category-based risk self-assessment architecture supervised directly by the regulator, with less involvement from the independent scientific community. This regulatory divergence foreshadows fertile ground for normative forum shopping by transnational platforms, which may choose to prioritize compliance in whichever jurisdiction imposes lighter transparency burdens toward third parties.

4.4. Looking ahead: toward an attention economy built on well-being rather than extreme emotional intensity

The horizon sketched out by this research and by regulatory responses points to a profound transformation of the digital attention economy. Source (4) describes the phenomenon of "social warming" as a gradual but real process of deterioration in our online and offline attitudes and behaviors — a vicious cycle of anger and outrage — that can nonetheless be corrected (4). The question is whether that correction will come through regulation, technological redesign, or a combination of both.

The DSA represents a significant first step, but its effective enforcement faces considerable challenges. Source (11) notes that early implementation reveals problems such as uneven compliance, uncertain technical standards, funding constraints, and restrictions on data-sharing for replication. Nevertheless, the DSA marks an inflection point: research on platforms no longer depends on corporate discretion, but is grounded in public-interest regulation (11). Researchers now play a central role in shaping evidence-based oversight of digital platforms in Europe (11).

Source (15) suggests that, for businesses and citizens alike, the lever is not to "switch off" social media, but to reduce the incentives for conflict through brand-safety policies, information hygiene, friction-by-design, and algorithmic transparency. Source (3) adds that algorithms do not merely show polarizing content to users — they also teach users to produce more of it. When a user posts moral outrage and is rewarded with likes and reshares, they learn to post more outrage next time (3). Breaking this learning cycle requires interventions that act on the feedback mechanisms themselves, not just on content.

Ultimately, the challenge is to reconceive the attention economy. Over the past decade, value has been measured in terms of emotional intensity and time-on-platform. An alternative model built on well-being, quality of debate, and informational accuracy would require not only technical changes, but also a realignment of economic incentives and greater public awareness of the mechanisms of emotional manipulation. As source (4) notes, social warming is real, but it can be corrected. The question is whether the correction will arrive in time, and with sufficient depth.

Bibliography

(0) Regulation (EU) 2022/2065 of the European Parliament and of the Council of 19 October 2022 on a Single Market For Digital Services and amending Directive 2000/31/EC (Digital Services Act). OJ L 277, 27.10.2022, p. 1–102. Available at: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32022R2065

(1) Ayolov, P. (2026). Rage as Revenue: How Anger Became the Currency of Digital Media. PhilArchive. Available at: https://philpapers.org/rec/AYORAR

(2) Cover, R. (2025). AI generation of rage bait: Implications for digital harms. New Media & Society. Published online 16 December 2025. DOI: 10.1177/14614448251400675. Available at: https://sage.cnpereading.com/doi/10.1177/14614448251400675

(3) Karavelioğlu, B. (2025). Rage Bait in the Digital Sports Ecosystem: A New Media Algorithm Manipulating Fan Engagement [orig. Turkish]. Spor Eğitimi ve Bilim Dergisi, Vol. 1, No. 2, 30 December 2025, pp. 76–90. Available at: https://dergipark.org.tr/tr/pub/sporegitimivebilimdergisi/article/1838693

(4) Arthur, C. (2021). Social Warming: The Dangerous and Polarising Effects of Social Media. Oneworld Publications / MIT Press. ISBN: 9780861543175.

(5) Brady, W. J., Doyle, M., Elnakouri, A., Finkel, E. J., Jackson, J. C., Kteily, N., Parker, V., Puryear, C., et al. (2026). Redesigning algorithms to intervene on social norm misperceptions during a national election. Nature. Published 27 May 2026. DOI: 10.1038/s41586-026-10536-1. Available at: https://www.nature.com/articles/s41586-026-10536-1

(6) McLoughlin, K., et al. (2024). Misinformation exploits outrage to spread online. Science. Published 29 November 2024.

(7) McAlaney, J. (2025). Rage bait: the psychology behind social media's angriest posts. The Conversation. Published 9 December 2025. Available at: https://theconversation.com/rage-bait-the-psychology-behind-social-medias-angriest-posts-271041

(8) Oxford University Press. (2025). The Oxford Word of the Year 2025 is rage bait. Available at: https://corp.oup.com/word-of-the-year-2025/

(9) Study: Social media probably can't be fixed. (2025). Ars Technica. Published August 2025.

(10) European Commission, Joint Research Centre (JRC). (2024). Digital Services Act: Application of the risk assessment framework for algorithmic recommender systems. Available at: https://digital-strategy.ec.europa.eu/en/library/digital-services-act-application-risk-assessment-framework-algorithmic-recommender-systems

(11) The DSA's investigator access provisions in practice: opening the black box of platforms? (2025). Verfassungsblog. DOI: 10.17176/20250307-110400-0. Available at: https://verfassungsblog.de/dsa-investigator-access/

(12) R, J. (2026). Scrolling Without Consent: Ragebait, Doomscrolling, and the Chaos Inside Our Heads. Jay R. Published 3 February 2026. ISBN: 9798224807574.

(13) Christin, A. (2025). Why we can't stop clicking on rage bait. Stanford Report. Published 2 December 2025. Available at: https://news.stanford.edu/stories/2025/12/rage-bait-explained-oxford-word-year

(14) Rose-Stockwell, T. (2023). Outrage Machine: How Tech Amplifies Discontent, Disrupts Democracy — and What We Can Do About It. Legacy Lit / Hachette Books, New York. First edition.

(15) Briggs, M. (2026). Outrage Machine: How Social Media Algorithms Learned to Profit from Human Anger. Heritage Books. ISBN: 9786266728124.

(16) Azmi, F. T. (2025). From 'Brain Rot' to 'Rage Bait': Stuck in the Online Loop. Outlook India. Published 14 December 2025. Available at: https://www.outlookindia.com/culture-society/from-brain-rot-to-rage-bait-stuck-in-the-online-loop

(17) Eastwood, M. W. (2025). The Anatomy of Outrage: The Proliferation of Rage Baiting. Independently published. Published 29 September 2025. ISBN-13: 979-8267712309.

(18) Guadagno, R. E. (2025). Virality, social validation, and emotional contagion on social media. In Psychological Processes in Social Media: Why We Click, Chapter 6, pp. 133–148. Elsevier / ScienceDirect. Published 2025.

(19) Nanji, N. (2025). 'Rage bait' named Oxford word of the year 2025. BBC News. Published 30 November 2025. Available at: https://www.bbc.com/news/articles/cewjxqvqzgyo

(20) Weiß, E.-M. (2025). Social media won't get any better without algorithms. Heise. Published 18 August 2025. Available at: https://www.heise.de/en/news/Social-media-won-t-get-any-better-without-algorithms-10539734.html

(21) Ofcom. Year 1 Online Safety Risk Assessments Report. Published 4 December 2025, under the Online Safety Act 2023 (United Kingdom). Available at: https://www.ofcom.org.uk/siteassets/resources/documents/online-safety/research-statistics-and-data/os-standards/online-safety-risk-assessments-report-year-one.pdf