references · lineage

reading list

The works the framework draws on, the works it argues against, and the works that did the formal heavy lifting. Grouped by what kind of crossing the work performs with the framework — not by canonical genre.

The framework's lineage

These are the critical-theoretic objects the framework is coupled to, not just synthesized from. Each does work specific to the closed-loop argument: Adorno on the form of capture; Marcuse on how compliance is manufactured as freedom; Stiegler on the technological substrate of psychic interiority; Han on the achievement-society's particular register of submission; Crary on the colonization of the body's slowest-leaking regime; Lyotard on the libidinal economy as historical-political object; Deleuze on the architectural shift from discipline to control; Mannheim on cohort formation re-formalized as a settled-form claim about a population.

Theodor Adorno, Minima Moralia (1951) — and the long arc through Dialectic of Enlightenment (with Horkheimer, 1944) and the culture-industry essays. The epigraph on the cover is from §50.
Herbert Marcuse, One-Dimensional Man (1964) — repressive desublimation, the production of freedom whose shape is the system's reproduction. The closed loop is the same dynamic with the substrate of the modeling exposed.
Bernard Stiegler, Technics and Time (3 vols., 1994–2001); Taking Care of Youth and the Generations (2010); States of Shock (2012). Tertiary retention, psycho-technology, the proletarianization of knowledge under digital platforms.
Byung-Chul Han, The Burnout Society (2010), The Transparency Society (2012), Psychopolitics (2014). The achievement-subject who exploits themselves; the tyranny of transparency as the platform's working condition.
Jonathan Crary, 24/7: Late Capitalism and the Ends of Sleep (2013), Scorched Earth (2022). Sleep as the last weakly-coupled regime. The wellness industry's colonization of it appears in the framework as the captured-resonance argument's late stage.
Jean-François Lyotard, Libidinal Economy (1974). The term in the framework's title is taken seriously rather than as flourish. Lyotard's argument that capital operates on libidinal investment, not just on surplus value, is the framework's starting position.
Gilles Deleuze, "Postscript on the Societies of Control" (1990). The dividual condition — the figure the closed-loop framework formalizes as the mutual predictability of the four registers under platform reading — is Deleuze's, sharpened.
Karl Mannheim, The Problem of Generations (1928). The framework's cohort-gradient claim is Mannheim re-formalized through the ergodic theory of learning systems: a population whose distributions converge under a common policy to a settled form, with developmental-stage-at-introduction as moderator.
Michel Foucault, Discipline and Punish (1975), The History of Sexuality I (1976), the Collège de France lectures on governmentality and biopolitics (1977–1979). The framework's "capture without coercion" is the contemporary form of what Foucault tracked across the disciplinary–biopolitical hinge.
Marshall McLuhan, Understanding Media (1964). The medium-as-message thesis becomes, on the framework's reading, a structural claim about how delivery-substrate shapes the substrate of subjectivity itself.

Cynicism and the captive

The Sincere Captives essay turns on Sloterdijk's diagnosis and Žižek's sharpening of it, against the background of Burroughs's language-as-virus.

Peter Sloterdijk, Critique of Cynical Reason (1983). The cynical professional is the framework's transitional figure; the sincere captive is what completed cynicism looks like at platform scale.
Slavoj Žižek, The Sublime Object of Ideology (1989). "They know very well, but still they do" — the formulation that lets the cynical position be read structurally rather than psychologically.
William S. Burroughs, The Naked Lunch (1959), The Soft Machine (1961), The Ticket That Exploded (1962). Language as virus, the cut-up method as compositional logic. The sincere captive's metabolized platform-grammar is Burroughs's Control completed.
Karl Marx, the chapter on commodity fetishism in Capital I (1867). The alienated worker who has somewhere to stand is the cynical professional's distant ancestor; the captive is what happens when the standing-position is foreclosed.

Democratic substrate and the pedagogical question

For The Citizen After the Pause: what the older democratic forms presupposed, and the pedagogical-political register the framework's prescriptions sit in.

Jürgen Habermas, The Structural Transformation of the Public Sphere (1962), Between Facts and Norms (1992). The deliberative democratic tradition the framework's structural diagnosis cuts against — not to dismiss its results, but to specify why repair pitched at the participant rather than the apparatus is structurally misaligned.
John Rawls, Political Liberalism (1993). The participant Rawls's framework presupposes is the older citizen; the recruitment-bias toward the partial-interiority-retained subgroup is the surface symptom of a structural absence the framework names.
Adam Phillips, On Kissing, Tickling, and Being Bored (1993), Becoming Freud (2014). The productive boredom the framework's pedagogical argument defends as a developmental resource.
Gaston Bachelard, The Poetics of Reverie (1960). Boredom as the attic of freedom; the slow-time domestic register the platform does not host.

The body and the affect-grammar

For the captured-resonance and somatic-optimization arguments. The framework is committed to the body as the differential against which extraction integrates — these are the works that ground that commitment.

Melanie Klein, Envy and Gratitude (1957) — object relations, paranoid-schizoid and depressive positions, projective identification.
D. W. Winnicott, Playing and Reality (1971) — transitional objects, potential space, the good-enough environment. What the platform overwrites when it forecloses the conditions of not-being-engaged.
Sigmund Freud, The Ego and the Id (1923) — the superego as the inheritor of the moral-religious archive. The metric superego is the platform's recension of the same internal voice.
Gilles Deleuze and Félix Guattari, Anti-Oedipus (1972) — the desiring-machine theorization the libidinal-economy reading sits inside.
Jean Baudrillard, Simulacra and Simulation (1981), Symbolic Exchange and Death (1976). The closed loop is not a simulation problem in Baudrillard's sense, but the operational architecture through which the dynamics Baudrillard tracked become structurally unavoidable.

The mathematical apparatus

For the formal results: §3 (the six instruments), §4 (foreclosure and reclamation), §5 (the three modes), §6 (cohort dynamics and mixing).

Søren Asmussen, Applied Probability and Queues (2nd ed., 2003). Chapter V (renewal theory) supplies the exponential survival bound the foreclosure lemma rests on.
Olav Kallenberg, Foundations of Modern Probability (2nd ed., 2002). §6.3 on regular conditional probabilities and disintegration — the measure-theoretic ground for "undefined-on-trajectory" in Theorem 1, Step 2.
Yoshiki Kuramoto, Chemical Oscillations, Waves, and Turbulence (1984). The coupled-oscillator model under which captured resonance is formalized (§3.6).
Alan G. Hawkes, "Spectra of some self-exciting and mutually exciting point processes" (Biometrika, 1971). The marked point-process formulation of somatic optimization (§3.5) is a multi-mark generalization of the Hawkes process.
Hassler Whitney, "Tangents to an analytic variety" (Annals of Mathematics, 1965). The stratification theorem invoked in the proof of Lemma 2(iii) to handle bifurcation along the regularization path.
Vladimir Vapnik, Statistical Learning Theory (1998); John Tsitsiklis & Benjamin Van Roy, the Markov-decision-process and policy-gradient literature broadly. The online-learning dynamics in equation (4) sit inside the standard policy-gradient framework.

Recent / contemporary

Works whose object overlaps with the framework's, and whose specific contributions the framework's diagnostic builds on or argues against.

Shoshana Zuboff, The Age of Surveillance Capitalism (2019). The extraction account the closed-loop framework specifies a structural mechanism for. Where Zuboff's framing remains at the level of rights-and-behavioral-modification, the closed-loop framework formalizes the architecture through which the modification operates.
Yanis Varoufakis, Technofeudalism (2023). The political-economic rendering of platform capital the framework supplies the libidinal counterpart to.
Kate Crawford, Atlas of AI (2021). The material substrate of platform inference — energy, labor, data — that the framework's formal apparatus deliberately leaves outside its scope.
Cory Doctorow, the enshittification essays (2023–). The political-economic decay-path of platforms as their incentive structure compounds; the closed-loop framework specifies a prior structural fact about the apparatus the decay operates on.
Whitney Phillips, This Is Why We Can't Have Nice Things (2015); the broader literature on platform-mediated discourse pathology (Marwick, boyd, Marantz, Tufekci). The empirical surface the framework's cohort-gradient prediction must engage with if it is to be tested.

Empirical backing for the framework's claims

A framework that specifies refusable structural claims needs to be read against the empirical literature that backs (or could refute) each claim. The works below are grouped by which framework claim they bear on. Citation density is uneven: some claims have a thick empirical literature, some have only suggestive findings, some have none yet. Where the framework's prediction is genuinely open, that is named below rather than papered over.

On the closed loop's architecture (Equations 1–6)

The platform-as-learning-system formulation in equation (4) — online gradient ascent on engagement reward — is the operative description of production-scale recommender systems, not a speculative one:

Paul Covington, Jay Adams & Emre Sargin, "Deep Neural Networks for YouTube Recommendations" (RecSys, 2016). The YouTube system specified end-to-end, with the engagement objective explicit. The framework's policy $\pi(y_{1:t}; \theta_t)$ is the abstraction of exactly this kind of architecture.
Heng-Tze Cheng et al., "Wide & Deep Learning for Recommender Systems" (DLRS, 2016). Google's reference architecture for the same problem.
Maxim Naumov et al., "Deep Learning Recommendation Model for Personalization and Ranking Systems" (Meta, arXiv 2019). The production architecture behind Facebook/Instagram feed ranking.
Smitha Milli, Luca Belli & Moritz Hardt, "From Optimizing Engagement to Measuring Value" (FAccT, 2021). Maps the engagement-objective optimization the framework formalizes onto the empirical reality at Twitter.
Eytan Bakshy, Solomon Messing & Lada Adamic, "Exposure to ideologically diverse news and opinion on Facebook" (Science, 2015). The platform-side measurement of what equation (3) determines: who sees what.

On (A2) — engagement reward monotone in event count

The empirical claim that platform reward functionals are non-decreasing in engagement-event count is supported by the disclosed objective functions of production systems:

Allison Stanger et al. on platform incentive structures, and the broader literature on click-through rate (CTR), watch-time, and dwell-time as the operative objectives. The Meta whistleblower disclosures (Frances Haugen, 2021) include internal documents that make the engagement-event monotonicity explicit at the policy level.
Cory Doctorow, the enshittification essays (2023–) document the same structural fact from the political-economic side: platforms' optimization pressure pushes the reward function toward engagement-count-maximizing forms.

On (A1) and (A4) — inter-arrival distributions and user responsiveness

Bounded coefficient of variation on inter-arrival times, and positive response probability after stimulus:

Albert-László Barabási, Bursts: The Hidden Pattern Behind Everything We Do (2010), and the underlying paper "The origin of bursts and heavy tails in human dynamics" (Nature, 2005). Human-activity inter-arrival distributions are heavy-tailed but have bounded second moment in operating regimes.
Adam Sadilek & Henry Kautz on smartphone-mediated activity-arrival distributions; the broader human-dynamics literature has dozens of empirical fits to renewal-type processes.
Larry V. Hedges and Christopher Schimel type meta-analyses of notification response rates: smartphone notifications produce response probabilities reliably bounded away from zero across populations (~80% within 5 minutes for younger users).

On §3.1 surplus extraction and per-register routing

The marked-point-process formulation and the four-register decomposition:

Alan G. Hawkes, "Spectra of some self-exciting and mutually exciting point processes" (Biometrika, 1971) — the original.
Marian-Andrei Rizoiu et al., "Hawkes Processes for Events in Social Media" (Frontiers in Big Data, 2018). Hawkes processes applied to clickstream and engagement data; the multi-mark generalization the framework uses is standard in this literature.
Junier Oliva, Barnabás Póczos & Jeff Schneider on multi-task user modeling; the broader recommender-system literature on shared user embeddings is the empirical version of the framework's cross-register coupling.
Eric Hehman et al. and the platform-segmentation literature for the empirical operationalization of the four registers (Subject/Citizen/Person/Consumer mapped to content categories).

On §3.3 libidinal routing — the dividual condition

The empirical claim that mutual information across registers fills the off-diagonal under prolonged exposure has only fragmentary backing — this is one of the framework's more aggressive empirical predictions:

Michal Kosinski, David Stillwell & Thore Graepel, "Private traits and attributes are predictable from digital records of human behavior" (PNAS, 2013). The original empirical demonstration that single-channel engagement traces are sufficient to predict cross-channel attributes — the dividual condition's first surface symptom.
Sandra Matz et al., "Psychological targeting as an effective approach to digital mass persuasion" (PNAS, 2017). Cross-register inference operationalized as persuasion-targeting.
Aral, Walker and the broader social-media-influence literature on cross-channel behavioral contagion.

On §3.4 metric superego — convergence under prolonged exposure

The convergence of self-evaluation toward the platform's reward distribution. Three named mechanisms in the math doc, three distinct literatures backing each:

Free-energy minimization (Friston-style):

Karl J. Friston, "The free-energy principle: a unified brain theory?" (Nature Reviews Neuroscience, 2010). The variational free-energy framework the math doc references as one of the three mechanisms producing convergence.
Andy Clark, Surfing Uncertainty (2016). The active-inference literature making the framework's claim about predictive-machinery alignment concrete.

Operant reinforcement (Rescorla-Wagner style):

Robert A. Rescorla & Allan R. Wagner, "A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and nonreinforcement" (in Classical Conditioning II, 1972). The classical formulation; the math doc's Proposition 2 derivation under operant reinforcement points here.
B. F. Skinner, Schedules of Reinforcement (with Ferster, 1957). The variable-ratio schedule literature; what makes the platform's intermittent reward structure operationally indistinguishable from a gambling schedule.
Adam Alter, Irresistible (2017). The popular synthesis specifically applied to platform engagement, with empirical citations to the schedule-of-reinforcement literature.

Social comparison (Festinger-style):

Leon Festinger, "A theory of social comparison processes" (Human Relations, 1954). The founding statement.
Jean Twenge, iGen (2017), Generations (2023), and the broader literature on social-media-induced social comparison and its mental-health correlates.
Melissa Hunt et al., "No more FOMO: Limiting social media decreases loneliness and depression" (Journal of Social and Clinical Psychology, 2018). Quasi-experimental backing for the social-comparison route to metric-superego formation.

On §3.5 somatic optimization

The Hawkes-style cross-excitation backing has been cited above. The somatic-as-differential claim has additional backing in:

Allen Frances, Wilhelm Hofmann and the broader self-regulation literature on attentional capture by external stimuli. Smartphone usage as the production of saccadic and tactile micro-events at high frequency is the empirical surface the framework formalizes.
Larry Rosen, Nancy Cheever & L. Mark Carrier, iDisorder (2012); the broader literature on phantom vibration, checking compulsions, and notification anticipation as somatic phenomena.

On §3.6 captured resonance — circadian and attentional entrainment

The Kuramoto-style coupling of biological clocks to platform delivery:

Yoshiki Kuramoto, Chemical Oscillations, Waves, and Turbulence (1984) — the formal apparatus.
Anne-Marie Chang et al., "Evening use of light-emitting eReaders negatively affects sleep" (PNAS, 2015). Direct empirical demonstration of platform-driven circadian disruption.
Mariana Figueiro et al. on screen-light suppression of melatonin; the broader chronobiology literature on artificial-light–circadian coupling.
Adam Hyland, Hannah K. Mathew on smartphone notification timing and sleep-onset disruption.
Mathias Basner & David F. Dinges, "Sleep duration and chronic sleep debt: not the same" (Sleep, 2018). The chrono-debt argument's empirical neighbor.

On §3.2 chrono-debt — the foreclosed pause-time

Time-use research is the relevant empirical surface; the developmental chrono-debt claim is where the framework's prediction is most aggressive:

American Time Use Survey (BLS, ongoing). Time-use distributions across cohorts; smartphone-mediated time is now substantial across all age groups.
Common Sense Media, "The Common Sense Census" (multiple years). Adolescent screen-time and time-use breakdowns.
Larry D. Rosen et al. on inter-stimulus intervals in smartphone use. Median checking intervals on the order of minutes, with the realized distribution skewing far shorter than the reflective threshold the framework's foreclosure argument requires.
Gloria Mark et al., Attention Span: A Groundbreaking Way to Restore Balance, Happiness and Productivity (2023). Attentional residence-time on a single task in knowledge-work settings — the empirical surface of the foreclosure lemma at micro scale.

On the cohort gradient (§6.1) — the framework's most refutable claim

The framework predicts that long-lag autocorrelation in user state is moderated by developmental-stage-at-introduction to the loop. This is the framework's strongest empirical commitment; the relevant literature is unsettled.

Jean M. Twenge, Jonathan Haidt et al., iGen (2017), The Anxious Generation (Haidt, 2024). The cohort claim made empirically; the framework's prediction is the structural mechanism that would explain the cohort effect if it holds. Note that Twenge and Haidt's claims are contested (Candice Odgers, Andrew Przybylski, Amy Orben). The framework's commitment is to the structural prediction; whether the empirical record bears it out is what makes the prediction refutable.
Amy Orben & Andrew K. Przybylski, "The association between adolescent well-being and digital technology use" (Nature Human Behaviour, 2019); "Screens, teens, and psychological well-being" (Psychological Science, 2019). The skeptical position; small effect sizes in cross-sectional data.
Eric Klopack, Steve Cole et al. on adolescent stress biomarkers and platform exposure.
Russell Viner et al. on smartphone introduction studies in longitudinal cohorts.

On §3.3 mobilization — the disruption pattern

The political-form prediction: rapid onset, rapid dispersion, no consolidation. Empirical:

Zeynep Tufekci, Twitter and Tear Gas: The Power and Fragility of Networked Protest (2017). The most direct empirical mapping of the framework's mobilization claim.
Manuel Castells, Networks of Outrage and Hope (2012). The earlier formulation.
Tarleton Gillespie, Custodians of the Internet (2018) on platform-mediated political infrastructure.
The empirical literature on Arab Spring, Occupy, Black Lives Matter, Hong Kong, Sri Lanka, Bangladesh — each separately, all consistent with the framework's structural prediction.

On the LinkedIn fanatic / sincere captive

The phenomenology in Sincere Captives sits in suggestive distance to several empirical surfaces, but the captive-as-population claim is a framework-prediction, not an established demographic fact:

Whitney Phillips & Ryan M. Milner, You Are Here (2021). Platform-mediated affect economies as a population-scale phenomenon.
Brooke Erin Duffy, (Not) Getting Paid to Do What You Love (2017). The professional self as platform-output, with the affective costs.
C. Thi Nguyen, "How Twitter gamifies communication" (in Applied Epistemology, 2021) and Games: Agency as Art (2020) on value-capture as the metric-superego mechanism at the discursive layer.

On (A6) — the regularity assumptions

The local-maximum-is-isolated-with-negative-definite-Hessian condition is generic in differentiable optimization. The relevant pointer is to Yurii Nesterov, Lectures on Convex Optimization (2nd ed., 2018) for the standard treatment, plus the policy-gradient convergence literature (Sergey Levine et al. for the practical RL side, Tsitsiklis & Van Roy for the foundational analysis cited above).

On Mode A's mixing-time claim (§6.2)

The framework predicts $T_{\mathrm{mix}}$ on the order of months to years for engagement-monetized policies. Empirical anchor:

The TikTok algorithm-change literature (e.g., Catherine Stewart's reporting on internal algorithm shifts) shows cohort-level distribution changes taking quarters, not weeks.
YouTube's deemphasis of borderline content (2019 announcement, multiple follow-up studies) — engagement-distribution shifts measured at multi-month timescales.
Standard Markov-chain mixing-time bounds (David Levin & Yuval Peres, Markov Chains and Mixing Times, 2017) give the formal apparatus.

On Mode B's metabolic boundedness

The framework's claim that maintenance labor is bounded by a budget constrained by exogenous social-reproduction conditions:

Silvia Federici, Wages Against Housework (1975), Revolution at Point Zero (2012). The social-reproduction literature the framework's metabolic-boundedness claim draws on.
Susan Folbre, The Invisible Heart (2001), Who Pays for the Kids? (1994). Time-poverty and care-work distribution.
Maria Mies, Patriarchy and Accumulation on a World Scale (1986). The classical formulation.
Kathi Weeks, The Problem with Work (2011). The contemporary synthesis.

On Mode C's information-theoretic bound

Blackwell-style sufficiency arguments for noisy-channel optimization:

David Blackwell, "Equivalent comparisons of experiments" (Annals of Mathematical Statistics, 1953). The classical statement.
Thomas Cover & Joy Thomas, Elements of Information Theory (2nd ed., 2006). Standard reference for the data-processing inequality the Mode C argument leans on.
Cynthia Dwork et al. on differential privacy as the operationalization of Mode C–style information bounds. The math is shared; the framing is different.

Where the empirical record is genuinely thin

To be specific about what the framework predicts but does not yet have strong empirical confirmation for:

The cohort gradient at the developmental level — open, contested literature.
The mutual-information convergence across the four registers — fragmentary evidence; needs purpose-built measurement.
Mode C's effectiveness at population scale (whether the cohort's multiple-legibilities practice in fact bounds platform inference) — no direct measurement exists to the author's knowledge.
The mixing-time prediction under architectural intervention — Mode A regulation experiments at scale (DSA/DMA effects) are still propagating.
The captive-as-population claim in Sincere Captives — phenomenological, with suggestive empirical neighbors but no rigorous population-level measurement.

These are the framework's most aggressive empirical commitments, and naming where the record is thin is part of what the framework specifies in its own scope statement (§7 of the math doc). They are not weaknesses to be hidden — they are the targets the framework specifies for whoever takes it up empirically.

The author's prior work

The Loot Loop (2017). The metropolitan-adolescent argument the Citizen After the Pause essay reframes through the closed-loop framework's cohort-gradient prediction.

A note on use

The framework does not require the reader to know these works to follow the argument. They are the substrate the author was thinking inside while writing. Where the prose names a specific argument from a specific thinker (Sloterdijk's cynical reason, Burroughs's language as virus, Mannheim's cohorts, Asmussen's renewal bound), the relevant work is the one to find. The rest is recommended on the strength of the framework's diagnosis: if the reading the framework is doing has purchase, then these are the works that did the work the framework's prose is building on.