Designed to Hook You: The Economy of Attention Explained

Introduction: From Human Attention to the Economic Architecture of Digital Engagement

Attention refers to a limited capacity to select stimuli, maintain goals, and allocate mental resources. In the contemporary economy, this capacity is no longer merely a psychological condition of consumption: it also functions as a unit of social exchange, public visibility, and economic accumulation (Heitmayer, 2024; Loewenstein & Wojtowicz, 2025). Digital platforms organize interfaces, recommendations, and notifications to prolong user engagement, because each interaction generates information about preferences and monetization opportunities.

The problem, therefore, is not simply that people use their devices a lot. It is that the technical architecture distributes attention through private, opaque rules geared toward interaction metrics. This distribution can affect self-control, autonomy, family relationships, and public deliberation (Benischke & Post, 2026; Jones, 2024). Examining addiction requires linking the psychology of attention to the economic model that converts observed behavior into revenue.

Attention as a Scarce Resource and Revenue for Digital Platforms

From Attention as a Symbolic Currency to Monetizable Behavioral Surplus

Economic literature presents attention as a scarce resource whose allocation explains decisions that deviate from classical assumptions of information and rationality (Loewenstein & Wojtowicz, 2025). On social media, this scarcity takes on a symbolic dimension: attention can be accumulated, exchanged, and converted into visibility for products, political messages, or research (Heitmayer, 2024). The platform does not necessarily sell a finished product; rather, it organizes user and advertiser access to a limited human capacity.

The model becomes more extractive when interactions are recorded, classified, and used to infer future behaviors. Behavioral surplus refers to data produced beyond what is necessary to provide a service, which can feed into profiles, predictions, and targeted advertising (Jones, 2024). The perception that an algorithmic profile is accurate is linked to a greater sense of surveillance and to attempts to modify advertising settings (Zhang et al., 2024). Thus, personalization can increase perceived convenience and, at the same time, reveal the asymmetry between those who generate data and those who decide how to exploit it.

How Algorithms Turn User Concentration into Market Power

Algorithmic attention rent arises when a platform controls the allocation of visibility and can extract value from users, providers, and advertisers (O’Reilly et al., 2024). The platform’s growth expands the amount of observable behavior and allows for fine-tuning of content selection, whereby user concentration reinforces intermediation capacity. This dynamic is not simply a matter of having many visitors: it involves governing what appears, when it appears, and what signals providers receive about demand.

Recommendation systems also shape supply. A simulation model using agents and reinforcement learning showed that different designs can skew providers’ perceptions of user preferences and alter the variety of content offered (Darani & Aghaie, 2025). Opacity makes it difficult to assess whether the selection favors satisfaction, diversity, or profitability. Therefore, the proposal to require information on operational metrics and the monetization of attention links economic transparency with algorithmic accountability (O’Reilly et al., 2024). Psychological scarcity thus becomes structural power, an issue that leads to an examination of specific mechanisms of capture.

Persuasive Designs and Algorithmic Recommendations: The Mechanisms Underpinning Continued Use

Personalization, content discovery, and engagement optimization

Personalization sustains continued use by reducing search effort and increasing the likelihood of finding content aligned with prior interests. In a mixed-methods study on TikTok, “like” rates, trending hashtags, and video duration influenced the likelihood of recommendation; interviews linked personalization to greater discovery and engagement, though also to homogeneity and echo chambers (Zhou, 2024). These associations do not prove that each component alone leads to a longer session, but they show how the system links behavioral signals to future visibility.

A comparison of short-form video platforms adds another layer. In a survey of college students, TikTok received higher ratings than Instagram Reels and YouTube Shorts for recommendation accuracy, serendipity, and perceived ease of use; on TikTok and Instagram Reels, these features were indirectly associated with addictive use through engagement (Roberts & David, 2025). The finding identifies a plausible chain of events, although the cross-sectional design does not establish a causal direction. Personalization can facilitate access and, at the same time, narrow the range of stimuli that the user comes to consider.

Intermittency, Automation, and Platforms as Hostile Attention Scaffolds

Platforms can be understood as attentional scaffolds: external structures that guide the selection and maintenance of attention. Voinea, Marin, and Vică compare them to slot machines because they combine intermittent rewards, emotionally arousing stimuli, and uncertain expectations about the next piece of content (Voinea et al., 2024). The analogy does not mean that every session leads to clinical addiction, but it does describe an architecture that makes it difficult to anticipate when a particularly rewarding stimulus will appear.

Engagement also operates through automation. When the interface strings videos together, reduces exit options, and selects emotionally charged stimuli, the user can shift from a deliberate decision to a habitual response (Voinea et al., 2024). Emotion-guided attention also extends to the collective level: moral outrage can spread within groups and generate shared attitudes (Voinea et al., 2024). In this sense, design not only organizes individual moments but also circuits of affective contagion.

The literature on persuasive design introduces a methodological caveat. A meta-analysis of 92 randomized controlled trials on mental health apps found clinical improvements compared to controls, with an effect size of g = 0.43, but found no association between the number of persuasive principles and engagement or efficacy (Valentine et al., 2025). This finding precludes treating all persuasion as harmful: the effect depends on the purpose, the metric, and the context of use. It also shows that “engagement” lacks a stable definition, as the studies used 25 different metrics and some did not report engagement data (Valentine et al., 2025).

From Guided Attention to Problematic Use of Short Videos

Problematic use arises when guided attention interacts with personal dispositions and automatic responses. In a survey of 561 Chinese participants, a serial mediation model indicated that automaticity and value-guided attention fully mediated the relationship between self-control and problematic use of short videos (Zhu & Fong, 2025). Individuals with lower self-control tended to respond more automatically and to selectively attend to valued stimuli, a process linked to problematic outcomes. The study validates a mechanism consistent with the I-PACE model, but its cross-sectional nature limits temporal inferences.

Evidence regarding platform characteristics suggests that precision, ease of use, and discovery can intensify engagement (Roberts & David, 2025). Even so, not all frequent interaction is problematic, and duration of use does not in itself equate to a loss of control. Personalization can reduce enjoyment when removed, as shown by a two-week experiment with 88 TikTok users: a less personalized feed reduced frequency and duration, increased self-regulation, and decreased enjoyment (Dekker et al., 2025). Persuasive design thus supports both perceived utility and a potential functional dependence.

When Capturing Attention Transforms the Individual Experience and the Public Sphere

Self-control, value-guided attention, and vulnerability to problematic use

The link between self-control and problematic use should not be interpreted as an exclusively individual responsibility. The available evidence describes a cognitive mediation process in which lower self-control is associated with automaticity and value-guided attention, which in turn are linked to problematic consumption of short videos (Zhu & Fong, 2025). The interface thus taps into a preexisting disposition and transforms it into a sequence of easily repeatable responses. Vulnerability depends on the interaction between individual capacities, valuable stimuli, and the architecture of choice.

Differences in attentional regulation are also not evenly distributed. The philosophical literature on digital distraction argues that variations in top-down and bottom-up regulation, internet engagement, and control over reward-seeking are related to socioeconomic factors (Kärki, 2024). When these differences affect classroom learning or tasks requiring complex cognition, distraction can threaten equal opportunity (Kärki, 2024). Interventions focused solely on willpower shift onto the user a cost that design helps to create.

Inequality in Digital Distraction and the Role of Family Relationships

Family experiences influence the risk of problematic adolescent use. A systematic review found consistent associations between positive parent-child relationships, a supportive family environment, and lower levels of problematic social media use (Vossen et al., 2024). Affectionate, responsive, and demanding parenting was associated with fewer problems, while inconsistency and control were associated with more problems (Vossen et al., 2024). In contrast, internet-specific mediation showed less consistent results: restrictive rules were associated with lower levels of problematic use, while reactive restrictions showed the opposite relationship.

Parental “phubbing” showed a consistent positive association with problematic adolescent use (Vossen et al., 2024). This finding shifts the explanation from technical monitoring toward the quality of everyday interaction. The review also notes that most studies were cross-sectional and used heterogeneous definitions of parenting, so a firm causal sequence cannot be established (Vossen et al., 2024). This attentional inequality is thus reproduced in households with varying resources for sustaining conversations, setting limits, and providing support.

Engagement, Satisfaction, and Algorithmic Amplification of Divisive Content

Optimizing interaction may prioritize content that generates activity without meeting the audience’s reflective preferences. A pre-registered algorithmic audit on Twitter found that, compared to reverse chronological order, interaction-based ranking amplified emotionally charged content that was hostile toward the outgroup (Milli et al., 2025). Users reported that these messages made them feel worse about the opposing political group, and they did not prefer the tweets selected by the algorithm (Milli et al., 2025). The revealed behavior—such as clicks, shares, or “likes”—thus serves as an imperfect indicator of well-being.

A ranking based on stated preferences reduced angry, partisan, and out-group-hostile content, although it could reinforce content consistent with prior attitudes (Milli et al., 2025). The solution is not to replace one metric with another without evaluating its distributive and epistemic effects. Recommendation systems also influence what providers interpret as demand, thereby altering variety and supply (Darani & Aghaie, 2025). The individual capture of attention can contribute to a more polarized public sphere—an issue that requires distinguishing between participation, satisfaction, and the quality of information.

Ambivalent Evidence: Engagement Effectiveness, User Well-being, and the Limits of Personalization

The controversy over whether personalization actually increases engagement

Recent experimental evidence supports a relationship between personalization and usage, but does not warrant an unlimited generalization. When 88 TikTok users were exposed to a less personalized feed for one week, the frequency and duration of use decreased, while self-regulation increased and enjoyment decreased (Dekker et al., 2025). The within-participant design and the collection of objective measures strengthen the comparison, although the two-week follow-up limits the assessment of persistent effects.

Other studies are correlational. A survey of 555 students linked perceived opportunities for recommendations, serendipity, and ease of use to engagement and, through this, to addictive use (Roberts & David, 2025). The survey does not reveal whether the platform produces these perceptions, whether users with a greater predisposition value them more, or whether both mechanisms operate simultaneously. The controversy stems, in part, from the difference between manipulating a flow and measuring retrospective evaluations.

Revealed Preferences versus Satisfaction, Autonomy, and Social Outcomes

Revealed preferences are useful for predicting clicks or shares, but they may deviate from stated satisfaction. The Twitter audit showed precisely that an engagement-oriented algorithm amplified content that users did not prefer and that worsened their assessment of the political outgroup (Milli et al., 2025). The platform may optimize a behavioral signal that the user, upon reflection, rejects. This gap calls into question the equivalence between spending more time on the platform and benefiting from the service.

Autonomy is also compromised when users know little about how recommendations are generated and have little control over their data. Interviews with young Finns showed that the systems limited opportunities for self-expression and required them to adapt to the system’s logic, although participants also used resources to obtain benefits for themselves (Jylhä et al., 2025). Personalization, therefore, combines practical utility with dependence on external rules. Evaluating it requires considering agency, satisfaction, and diversity—not just duration.

What Experiments, Surveys, Interviews, and Simulation Models Contribute

Each method sheds light on a different aspect of the phenomenon. Experiments identify imminent changes in behavior and experience; surveys compare perceptions and associations across platforms; interviews reconstruct conceptions of agency; and simulations explore how algorithmic rules modify providers and supply (Darani & Aghaie, 2025; Dekker et al., 2025; Jylhä et al., 2025; Roberts & David, 2025). No single method, on its own, establishes long-term social effects.

A meta-analysis of mental health apps adds a caveat regarding measurement: 25 different engagement indicators prevented a coherent overall assessment (Valentine et al., 2025). Future research must distinguish between exposure, interaction, enjoyment, self-regulation, satisfaction, and problematic use. Such methodological precision paves the way for a governance agenda that does not confuse commercial effectiveness with well-being.

Research Gaps and Alternatives for Governing Digital Care

From individual responsibility to responsible design of recommendation systems

A focus on self-control and individual habits proves insufficient when interfaces incorporate attention-grabbing features, intermittent rewards, and automated selection (Roffarello et al., 2025; Voinea et al., 2024). Digital attention heuristics propose designing interfaces that respect autonomy, competence, and social relationships, rather than requiring users to exercise constant self-regulation (Roffarello et al., 2025). This approach shifts prevention from ex post blocking to the initial conditions of interaction.

Governance must also require transparency regarding metrics, attention allocation, and monetization (O’Reilly et al., 2024). Systems should be evaluated based on their effects on users and providers, because recommendations can simultaneously alter behavior, supply, and visibility (Darani & Aghaie, 2025). Design accountability does not mean eliminating all personalization, but rather justifying its objectives and making its consequences auditable.

Consumer justice, youth agency, and epistemic well-being as evaluation criteria

Consumer justice requires defining what constitutes fair treatment in a recommendation. Systematic research on equity distinguishes multiple conceptions of discrimination and classifies metrics for evaluating biases against users (Vassøy & A, 2024). No single metric is sufficient: the evaluation must consider access, exposure, quality, diversity, and potential harms. Youth agency adds the capacity to understand and negotiate the use of data, not just the formal possibility of accepting or rejecting a recommendation (Jylhä et al., 2025).

Epistemic well-being offers a criterion for assessing whether communication infrastructures maintain the conditions necessary to exercise agency in the public sphere (Hyzen et al., 2025). This framework connects datafication, algorithmization, and platformization with the production, circulation, and acquisition of knowledge. Together with consumer justice, it allows for the evaluation of systems based on their effects on informational autonomy, diversity, and judgment, in addition to their commercial metrics.

Open Questions Regarding Causality, Long-Term Effects, and Differences Between Platforms

Doubts persist regarding the duration of effects, the direction of causality, and comparability across services. The TikTok experiment observed changes over two weeks, while several surveys measured associations at a single point in time (Dekker et al., 2025; Roberts & David, 2025). The existing literature also relies heavily on cross-sectional designs (Vossen et al., 2024). Long-term follow-ups, verifiable behavioral measures, and designs that disentangle user selection, interface features, and recommendation effects are needed.

The differences between TikTok, Instagram Reels, YouTube Shorts, and Twitter preclude the adoption of a one-size-fits-all framework. Objectives, formats, interaction signals, and the relationships among users, creators, and advertisers vary (Milli et al., 2025; Roberts & David, 2025). Research must compare platforms using common metrics, record stated and revealed preferences, and incorporate social and epistemic outcomes. Only then can governance address specific mechanisms rather than regulating an abstraction called “social media.”

Conclusion: Regaining attention requires questioning the business model and redesigning its incentives

The digital economy transforms a limited human capacity into a source of prediction, visibility, and revenue. Personalization, recommendations, and intermittent rewards can sustain continued use, but their effects depend on autonomy, satisfaction, vulnerability, and social context (Dekker et al., 2025; Voinea et al., 2024). The discrepancy between observed interaction and self-reported well-being demonstrates that engagement is not sufficient as a measure of success (Milli et al., 2025).

Regaining attention requires modifying incentives that reward retention without distinguishing between valuable use and compulsive consumption. This implies transparency regarding the allocation and monetization of attention, equity audits, metrics of epistemic well-being, and designs that respect agency (Hyzen et al., 2025; O’Reilly et al., 2024; Vassøy & A, 2024). Individual responsibility still has a role to play, but it cannot replace the transformation of the architectures that make it profitable to distract, profile, and retain users.

References

• Benischke, M. H., & Post, C. A. (2026). From platform design to attention governance: Rethinking social media externalities. Journal of Management Studies, 63(7), 4358–4376. https://doi.org/10.1111/joms.70118
• Darani, M. M., & Aghaie, S. (2025). The impact of recommender systems on a platform’s content and outcomes: the role of providers and algorithm designs. Journal of Research in Interactive Marketing, 19(6), 917–935. https://doi.org/10.1108/jrim-04-2024-0198
• Dekker, C. A., Baumgartner, S. E., & Sumter, S. R. (2025). “For you” vs. “for everyone”: The effectiveness of algorithmic personalization in driving social media engagement. Telematics and Informatics, 101, 102300. https://doi.org/10.1016/j.tele.2025.102300
• Heitmayer, M. (2024). The second wave of attention economics: Attention as a universal symbolic currency on social media and beyond. Interacting with Computers, 37(1), 18–29. https://doi.org/10.1093/iwc/iwae035
• Hyzen, A., Van den Bulck, H., Puppis, M., Kulig, M., & Paulussen, S. (2025). Epistemic Welfare and Algorithmic Recommender Systems: Overcoming the Epistemic Crisis in the Digitalized Public Sphere. Communication Theory, 36(1), 46–57. https://doi.org/10.1093/ct/qtaf018
• Jones, J. (2024). Don’t fear artificial intelligence, question the business model: How surveillance capitalists use media to invade privacy, disrupt moral autonomy, and harm democracy. Journal of Communication Inquiry, 49(1), 6–26. https://doi.org/10.1177/01968599241235209
• Jylhä, V., Hirvonen, N., & Haider, J. (2025). Algorithmic recommendations in the everyday life of young people: imaginaries of agency and resources. Information Communication & Society, 28(15), 2641–2657. https://doi.org/10.1080/1369118x.2025.2470227
• Kärki, K. (2024). Digital distraction, attention regulation, and inequality. Philosophy & Technology, 37(1). https://doi.org/10.1007/s13347-024-00698-z
• Loewenstein, G. F., & Wojtowicz, Z. (2025). The economics of attention. Journal of Economic Literature, 63(3), 1038–1089. https://doi.org/10.1257/jel.20241665
• Milli, S., Carroll, M., Wang, Y., Pandey, S., Zhao, S. Y., & Dragan, A. D. (2025). Engagement, user satisfaction, and the amplification of divisive content on social media. PNAS Nexus, 4(3), pgaf062. https://doi.org/10.1093/pnasnexus/pgaf062
• O’Reilly, T., Strauss, I., & Mazzucato, M. (2024). Algorithmic attention rents: A theory of digital platform market power. Data & Policy, 6. https://doi.org/10.1017/dap.2024.1
• Roberts, J. A., & David, M. E. (2025). Technology affordances, social media engagement, and social media addiction: An investigation of TikTok, Instagram Reels, and YouTube Shorts. Cyberpsychology, Behavior, and Social Networking, 28(5), 318–325. https://doi.org/10.1089/cyber.2024.0338
• Roffarello, A. M., De Russis, L., & Lukoff, K. (2025). The digital attention heuristics: Supporting the user’s attention by design. ACM Transactions on Computer-Human Interaction, 32(4), 1–41. https://doi.org/10.1145/3725215
• Valentine, L., Hinton, J. D. X., Bajaj, K., Boyd, L., O’Sullivan, S., Sorenson, R. P., Bell, I. H., Vega, M. S., Liu, P., Peters, W., Mangelsdorf, S. N., Wren, T. W., Moller, C. I., Cross, S. P. M., McEnery, C., Bendall, S., Nicholas, J., & Álvarez‐Jiménez, M. (2025). A meta-analysis of persuasive design, engagement, and efficacy in 92 RCTs of mental health apps. NPJ Digital Medicine, 8(1), 229. https://doi.org/10.1038/s41746-025-01567-5
• Vassøy, B., & A, H. L. (2024). Consumer-side fairness in recommender systems: a systematic survey of methods and evaluation. Artificial Intelligence Review, 57(4). https://doi.org/10.1007/s10462-023-10663-5
• Voinea, C., Marin, L., & Vică, C. (2024). Digital slot machines: Social media platforms as attentional scaffolds. Topoi, 43(3), 685–695. https://doi.org/10.1007/s11245-024-10031-0
• Vossen, H. G. M., van den Eijnden, R. J. J. M., Visser, I., & Koning, I. M. (2024). Parenting and problematic social media use: A systematic review. Current Addiction Reports, 11(3), 511–527. https://doi.org/10.1007/s40429-024-00559-x
• Zhang, D., Strycharz, J., Boerman, S. C., Araujo, T., & Voorveld, H. A. M. (2024). Google knows me too well! Coping with perceived surveillance in an algorithmic profiling context. Computers in Human Behavior, 165, 108536. https://doi.org/10.1016/j.chb.2024.108536
• Zhou, R. (2024). Understanding the impact of TikTok’s recommendation algorithm on user engagement. International Journal of Computer Science and Information Technology, 3(2), 201–208. https://doi.org/10.62051/ijcsit.v3n2.24
• Zhu, J., & Fong, L. H. N. (2025). Self-control and problematic short-form video usage: the mediating roles of automaticity and value-driven attention. Behaviour and Information Technology, 44(14), 3609–3619. https://doi.org/10.1080/0144929x.2025.2452367