The Dilemma of the Attention Commons: when everyone performs, no one rests (Part 2 of 5)

In the previous text I proposed that contemporary vitrinization should be understood as an instance of the political economy of cognition at scale – and I opened the discussion through the cognitive layer, where framing acts on the individual. I showed how the framing effect, established by Tversky and Kahneman in 1981, has been industrialized by recommendation algorithms, which frame content at personalized scale for billions of users simultaneously.

The cognitive layer, however, explains how framing affects an individual decision. It does not explain why, even knowing we are being framed, even knowing the effects of networks on our mental health, even after reading Han and Türcke, we keep posting, scrolling, comparing. Why is this equilibrium stable?

The answer requires another toolkit: game theory. And it also requires recovering an earlier theoretical lineage – a tradition of thought about collective behavior that long predates contemporary discourse on social networks and that illuminates, with surprising relevance, part of what is at stake.

The lineage of crowds: from Le Bon to Türcke

In 1895, Gustave Le Bon published Psychologie des foules, known in English as The Crowd: A Study of the Popular Mind. The book is controversial – it has politically uncomfortable readings, it was mobilized by mass-propaganda theorists of suspect provenance, and it contains dated claims about race and gender that warrant no defense. Yet its analytical core is disconcertingly current: individuals immersed in a crowd display behavior they would not display in isolation. There is heightened suggestibility, affective contagion, a temporary dissolution of individual judgment in favor of a “collective soul” (his expression) that operates with a logic of its own.

Le Bon was describing physical crowds – mobs, demonstrations, audiences. But the mechanisms he identified – emotional contagion, suspension of critical reflection, formation of a transient psychic unity – anticipate with striking precision what we observe in digital crowds: virals, cancellations, waves of outrage or euphoria that cross platforms in hours and shape individual behavior at mass scale.

The lineage matters because it gives historical depth to the argument. Freud worked explicitly with Le Bon in Group Psychology and the Analysis of the Ego, of 1921, and incorporated him into what would become his reflection on the collective superego in Civilization and Its Discontents. Christoph Türcke, in Erregte Gesellschaft (Excited Society), retakes the problem of contagion in the form of the dispute for attention mediated microelectronically. Byung-Chul Han, in In the Swarm: Digital Prospects, describes the digital collective precisely as a crowd without a body – an aggregation of isolated individuals whose simultaneity in digital space produces effects analogous to those of the physical crowd, but without the counterpart that the crowd once had: presence, the body, the possibility of deliberation. It is a continuous tradition, with a hundred and thirty years of depth, and one that deserves to be acknowledged as such.

The game-theoretic question, however, operates in a different register. Le Bon describes what happens. Game theory asks why this equilibrium is stable.

Vitrinization as a Nash equilibrium of mutual exhaustion

The Hanian reading – burnout society as the self-exploitation of the performance subject – is phenomenologically powerful, but insufficient to model the structure of incentives that produces and sustains the symptom. The game-theoretic question is different: given the incentive configuration of the platform and the strategy of other users, what is the best response of a rational subject?

When we formulate it that way, the answer appears. The vitrinization equilibrium is stable because it is a dilemma of the commons, structurally analogous – though not identical – to the tragedy described by Garrett Hardin (1968) for natural common-pool resources (formally, vitrinization is closer to a coordination game with multiple stable equilibria, with path-dependence over which basin the system has fallen into, than to Hardin’s single-equilibrium tragedy – a distinction that strengthens, further on, the relevance of tipping points). Each individual, given the behavior of others, has incentive to perform. If no one posts, posting yields high visibility returns. If everyone posts, not posting costs invisibility. Since visibility became capital – a cultural transformation Paula Sibilia mapped in O Show do Eu (2008), generalizing Bourdieu – invisibility is symbolic disoccupation, failure in the regime that defines who exists.

The result is a Nash equilibrium: no different unilateral strategy improves the outcome of the one who deviates. Everyone performs, everyone exhausts themselves, no one can defect in isolation without cost. And as in any dilemma of the commons, the equilibrium is Pareto-inferior: we would all be better off if no one performed, but from any state of low performance, there are individual gains in deviating, and the high-performance equilibrium reasserts itself.

Schelling: rational microchoices, absurd macro-outcomes

Thomas Schelling, in Micromotives and Macrobehavior (1978), showed that undesired collective behaviors can emerge from perfectly rational microchoices. The canonical example is the segregation model: individuals with moderate preferences about the kind of neighborhood they want can produce, in aggregate, extreme racial segregation. There is no racism at the micro-scale that explains the macro outcome; it is the dynamic of the game that produces the outcome.

A methodological note worth registering. Schelling’s segregation model is, in retrospect, one of the earliest agent-based models (ABM) in the social sciences. Before computation made the methodology scalable, Schelling ran simulations on physical boards. Today, ABM is one of the principal tools for investigating the emergence of collective phenomena from local rules of heterogeneous individual behavior. It is precisely the toolkit I am developing, in another register, in my doctoral thesis – where I investigate the conditions under which cooperation emerges among cognitively heterogeneous agents. Vitrinization has the same formal structure as what ABM models well: heterogeneous agents, local rules of interaction, emergent global properties. I will return to this point in the last text of the series.

Vitrinization thus has the same logical structure as Schelling’s model. No user individually wants “to perform exhaustively for the approval of others”. What each wants is to maintain connection, avoid invisibility, feel relevant. But the aggregation of these microchoices, under the specific architecture of current platforms, produces a macro-condition that is the burnout society.

Schelling offers another useful contribution: the notion of tipping point. Social equilibria usually have more than one stable point, with unstable points between them. If we manage to perturb the system enough, we can transit to another equilibrium. But small perturbations are absorbed: the system returns to its prior equilibrium. This is precisely why individualized “digital-detox” campaigns typically fail – and why regulatory measures must be structural to have effect.

Endogenous preferences: the missing ingredient

There is an additional sophistication that my training leads me to notice, and that few texts on social networks incorporate adequately. In classical economic models, an agent’s preferences are exogenous – they enter the game with them, and the game does not change them. In the case of vitrinization, this is false.

Leon Festinger, in 1954, proposed the theory of social comparison: people evaluate their own conditions, opinions, and abilities in relation to those of others, especially in the absence of objective criteria. Decades later, George Akerlof and Rachel Kranton formalized this intuition in identity economics (starting in 2000, with systematic development in their book Identity Economics, 2010): my utility depends not only on what I have, but on how I stand relative to a reference group.

When we apply this to digital space, something important changes. My utility from posting a vacation photo depends not just on the photo; it depends on how it compares to the photos I see from others. If everyone posts spectacular vacations, posting an ordinary vacation reduces my utility – I look worse. To maintain constant utility, I need to post ever better vacations. But everyone faces the same incentive. The result is a Red Queen race: we run ever faster to stay in the same place.

This means individual preferences are endogenous to the very game. What I desire depends on what others desire, which depends on what I desire, in an infinite loop. For a game theorist, this is an interesting technical problem and still little explored in its application to platforms.

Where GMCR can contribute and where it meets its limit

The Graph Model for Conflict Resolution, developed by Keith Hipel, Marc Kilgour and team from the 1990s onward, is a formal structure for modeling conflicts with multiple decision-makers, options, states, and preferences. Its most recent extensions incorporate preferences under uncertainty, hypergraph-type preferences, and, more recently, preferences under social influence.

Imagining an application of GMCR to the “dilemma of the attention commons” is an intellectually provocative exercise: the actors would be (i) heterogeneous users, (ii) the platform as meta-player designing the architecture, and (iii) advertisers; options and states would refer to levels of performance, exposure, and engagement; preferences would be socially endogenous. The stability concepts of GMCR – Nash, GMR, SMR, SEQ – would map which states are robust to which perturbations.

Here, however, we encounter a first-magnitude methodological obstacle that deserves to be said without hedge. In GMCR, the elicitation of decision-makers’ preferences is the hardest step of the entire application of the method. It is not a technical difficulty to be circumvented: it is the central difficulty. Even in conflicts with few actors and few options – like the Cocó viaducts conflict I studied in my master’s – obtaining complete, consistent, and stable orderings of preferences over states is arduous work. Decision-makers rarely manage to articulate complete preferences. When they do articulate, they do so with internal inconsistencies, with context dependence in the interview, and – closing a recursion worth insisting on – with sensitivity to the framing of the question itself. In a “conflict” with billions of users, thousands of platforms, and dozens of regulatory jurisdictions, direct elicitation becomes unfeasible.

Possible paths around this difficulty include: (i) revealed preferences from behavioral data, with the known pitfalls of inferring preferences from behavior under manipulated framing; (ii) stylized models with postulated preferences, validated by their ability to reproduce observed phenomena; (iii) aggregation in representative classes rather than agent-by-agent modeling; (iv) ABM approaches, in which heterogeneous preferences are parameterized and equilibrium emerges by simulation. Each path has its own analytical cost.

Recognizing this difficulty is not admitting failure; it is honestly locating where the work to be done lies. There may be, indeed, an original contribution in the very formulation of elicitation protocols that recognize the framing recursion – the measurement instrument subject to the same bias it studies – as part of the method, rather than as a source of noise to be eliminated.

There is close contemporary work. Daron Acemoglu, with Asuman Ozdaglar and James Siderius, has since 2021 been publishing a series of papers formally modeling platforms, engagement, information propagation, and degraded equilibria. It is direct neighborhood – though with emphasis on propagation dynamics, while a contribution via GMCR or ABM would focus on multi-equilibrium stability analysis and on the modes of transition between basins of attraction.

The opening question

If vitrinization is a Nash equilibrium of mutual exhaustion, sustained by socially endogenous preferences, then two conclusions impose themselves.

First: individual solutions are, by construction, insufficient. To ask people to use the networks less is to ask each one to deviate unilaterally from a stable equilibrium. There may be individual heroes, but they pay the price of invisibility. The equilibrium reasserts itself. Digital-detox campaigns, screen-time apps, retreats without phones – all act at the individual margin, not on the structure of the game. They are moral palliatives for a systemic problem.

Second: if the equilibrium is stable given the architecture, changing the equilibrium requires changing the architecture. But the architecture – algorithms, metrics, interface design, business models – is not a given of nature. It is the choice of a small set of corporate actors, in specific regulatory environments. In other words: the architecture is political.

That is the topic of the next text.

---

References

ACEMOGLU, D.; OZDAGLAR, A.; SIDERIUS, J. “A Model of Online Misinformation”. Review of Economic Studies, 2024.

AKERLOF, G. A.; KRANTON, R. E. Identity Economics: How Our Identities Shape Our Work, Wages, and Well-Being. Princeton: Princeton University Press, 2010.

CASTRO, D. P. Efeito de Enquadramento no Modelo de Grafos para Resolução de Conflitos com uma Aplicação ao Conflito das Obras de Construção dos Viadutos do Cocó [Framing Effect in the Graph Model for Conflict Resolution with an Application to the Cocó Viaducts Construction Works]. Master’s thesis – Federal University of Ceará, Fortaleza, 2022.

FESTINGER, L. “A Theory of Social Comparison Processes”. Human Relations, vol. 7, no. 2, pp. 117-140, 1954.

FREUD, S. Group Psychology and the Analysis of the Ego. New York: W. W. Norton, 1990. [Original German: 1921.]

HAN, B.-C. In the Swarm: Digital Prospects. Cambridge, MA: MIT Press, 2017.

HARDIN, G. “The Tragedy of the Commons”. Science, vol. 162, no. 3859, pp. 1243-1248, 1968.

HIPEL, K. W.; KILGOUR, D. M.; FANG, L. Conflict Resolution Using the Graph Model: Strategic Interactions in Competition and Cooperation. Cham: Springer, 2018.

LE BON, G. The Crowd: A Study of the Popular Mind. Mineola: Dover, 2002. [Original French: 1895.]

SCHELLING, T. C. Micromotives and Macrobehavior. New York: W. W. Norton, 1978.

SIBILIA, P. O Show do Eu: a intimidade como espetáculo [The Spectacle of the Self: intimacy as spectacle]. Rio de Janeiro: Nova Fronteira, 2008.

TÜRCKE, C. Erregte Gesellschaft: Philosophie der Sensation. Munich: C. H. Beck, 2002.

TVERSKY, A.; KAHNEMAN, D. “The Framing of Decisions and the Psychology of Choice”. Science, vol. 211, no. 4481, pp. 453-458, 1981.