There is a body of knowledge here. Nobody owns it. It is split between the people who teach how to build the loop and the people who teach how to see it, and the two sides barely share a room. That split is the story.
The literature divides cleanly on a question it rarely states plainly: not whether operators build instrumented feedback loops to steer behaviour — that is uncontested — but whether those loops reliably produce, at population scale, the changes their operators intend. On that question the evidence is asymmetric. It is strong and experimental where the intervention is a discrete interface choice, and genuinely unsettled where the intervention is an engagement-ranked feed acting on something as diffuse as adolescent well-being. A faithful review has to hold both findings at once.
The literature strongly supports that the apparatus exists and is engineered to modify behaviour; strongly supports that it works in controlled interface experiments; but shows no settled consensus on the effect size of engagement-ranked feeds on population-scale well-being.
Ask where you would go to study the loop — the instrumented feedback cycle the three case studies describe — and you find there is no single door. There is no Department of Behavioural Modification at Scale. The knowledge exists, it is taught, it is examined and credentialed and sold; but it lives in a seam between two pedagogies that rarely meet.
One pedagogy teaches how to build the loop, and teaches it as a concentrated, confident, immediately employable skill. The other teaches how to see the loop, and teaches it scattered across a dozen departments that use different vocabularies and publish in different rooms. The operators learned to work the seam.
Below is the map as the research returned it — with the honest note that the build side verified far more cleanly than the critique side, which is itself a small instance of the asymmetry.
The discipline of building the loop has names, addresses, and syllabi. Its oldest academic home is persuasive technology: B. J. Fogg's lab at Stanford, founded in 1998 as the Persuasive Technology / Captology Lab and renamed the Behavior Design Lab in 2018. Its stated mission is openly pedagogical — to "teach good people how human behavior works so they can create solutions" — and Fogg's Persuasive Technology: Using Computers to Change What We Think and Do (Morgan Kaufmann, 2003) seeded an entire applied field, including the annual PERSUASIVE conference (from 2006).
From there the pedagogy runs straight into the business and design schools. Nir Eyal — whose "Hook Model" (trigger → action → variable reward → investment) is the canonical recipe for an engagement loop — taught it at the Stanford Graduate School of Business and the Hasso Plattner Institute of Design (the d.school). The loop is not a critique object here. It is the deliverable.
The applied-computing wing is just as concrete. Carnegie Mellon's Human-Computer Interaction Institute grants a named professional degree — the Master of Human-Computer Interaction — and the recommender-systems community has its own dedicated ACM venue, RecSys (first held in Minneapolis, 2007; roughly sixty percent industry attendance). These are real credentials in real buildings.
The most revealing strand is online controlled experimentation — A/B testing, the engine that actually tunes the loop. Its definitive text is Trustworthy Online Controlled Experiments by Ron Kohavi, Diane Tang and Ya Xu (Cambridge University Press, 2020), explicitly "based on practical experiences at companies that each run more than twenty thousand controlled experiments a year." Kohavi's own KDD work documents Bing running more than two hundred concurrent experiments a day across roughly a hundred million monthly users, with ninety percent of eligible users enrolled in some experiment at any moment.
And where do you go to learn it? Kohavi — formerly a Microsoft Technical Fellow and Corporate Vice President of Analysis and Experimentation — does not hold a professorship. He teaches A/B testing through paid cohort courses on the practitioner marketplace Maven, to over a thousand working professionals since 2021. The build-side curriculum has, in places, exited the academy altogether and become a product. That is the cleanest single fact in this whole map: the knowledge to construct the loop is now teachable, profitably, in a few evenings.
The knowledge to analyse the loop has no comparable address. It is distributed across the traditional academy, one face of the machine per discipline:
Science & Technology Studies supplies the frame that technology is socially shaped, not neutral. Critical data and algorithm studies — the newest cluster — lives mostly in standalone institutes rather than departments: the Oxford Internet Institute; Princeton's Center for Information Technology Policy; Data & Society; the AI Now Institute; Harvard's Berkman Klein Center; Yale's Information Society Project; Georgetown's Center on Privacy & Technology. Media and communications carries the political-economy strand (Couldry; Zuboff). Law carries dark patterns, privacy, and platform regulation. Sociology of work carries algorithmic management; anthropology, the ethnography of addictive design; public policy, the welfare and state-systems strand. And psychology and behavioural economics — Skinner's behaviourism, the Kahneman–Thaler–Sunstein "nudge," Cialdini's persuasion — sit underneath as the shared substrate both sides draw from.
The verified anchor in this pass is the Princeton CITP "Dark Patterns at Scale" study (Mathur, Mayer, Narayanan et al., 2019), which operationalises the construct precisely: interface choices "that benefit an online service by coercing, steering, or deceiving users into making unintended … decisions," found in nearly two thousand instances across eleven thousand shopping sites. Note where it was published: not at a critique venue, but at the HCI conference CSCW — a rare crossing of the seam.
The clearest evidence that these are two communities and not one is where they publish. The build side lives at the computing and data-mining conferences — KDD, WSDM, CIKM, ICSE, RecSys, CHI, CSCW — and, more than anywhere, inside corporate research that is never submitted to an external venue at all. Kohavi's own publication record runs through KDD, WSDM, ICSE, CIKM — all computing, none of them critique venues.
The critique side publishes at FAccT (the ACM Conference on Fairness, Accountability, and Transparency, founded only in 2018), AoIR, 4S, and in the law reviews. FAccT is the nearest thing to a shared room — but it is young, and it leans hard to the analysis side. The asymmetry is structural: the people building the loop and the people studying it are, for the most part, not in the same conversation, and have not been for the twenty years it took the loop to mature.
A coherent home may be forming, but late and in fragments. "Critical data studies" now exists as a named term and, in places, a program (Purdue runs one). A "Trust & Safety" curriculum has begun to be codified by a professional association rather than emerging from a discipline. Reading lists — most famously Tarleton Gillespie and colleagues' "critical algorithm studies" bibliography — function as de facto syllabi precisely because no settled canon yet exists.
The dates tell the story. The persuasive-technology textbook is from 2003; the build-side experimentation canon, 2020; the critique-side's nearest shared conference, 2018; the named "studies" programs, more recent still. The thing being studied is older than the field assembled to study it. The loop had a twenty-year head start.
This is not an academic curiosity. It is part of the paper's central claim about why public discourse runs a decade behind the operators. The knowledge required to build a behavioural loop is concentrated, vocational, and confidently taught — you can buy the course. The knowledge required to see one is spread thin across disciplines that don't share a vocabulary, a conference, or a degree, and none of which is accountable for the whole machine.
An asymmetry of pedagogy becomes an asymmetry of power. One side graduates practitioners who can ship a loop next quarter; the other graduates critics who can describe one face of it in a journal three fields away from the engineer. The loop wins the seam because the seam is where no one is responsible. Naming the hand — the editorial demand of the whole paper — requires first building the discipline that could name it.
The field's conceptual scaffolding is older than the platforms. Its load-bearing move is to treat instrumented design as deliberate behavioural engineering rather than neutral tooling. The behaviourist root is B. F. Skinner's work on operant conditioning, and in particular the finding that variable-ratio reinforcement — reward delivered on an unpredictable number of responses — produces the highest, most extinction-resistant response rates. That is the slot-machine schedule, and it reappears, undisguised, in the design literature on feeds.
Natasha Dow Schüll's Addiction by Design: Machine Gambling in Las Vegas (Princeton University Press, 2012) is the canonical bridge from that tradition to machine-mediated design: a roughly fifteen-year ethnography of how slot-machine design engineers a continuous "machine zone" of play. It remains the most-cited demonstration that addictive engagement is a design target, not a side effect.
The persuasion strand runs through B. J. Fogg's Persuasive Technology (Morgan Kaufmann, 2003), which founded "captology" — the study of computers as persuasive technology — and its trade descendant, Nir Eyal's Hooked: How to Build Habit-Forming Products (Portfolio, 2014), whose "Hook Model" (trigger → action → variable reward → investment) reads in retrospect as much a confession of intent as an instruction manual.
The political-economy strand reframes all of this as an extractive system. Shoshana Zuboff's The Age of Surveillance Capitalism (PublicAffairs, 2019) names behavioural surplus — the exhaust of user behaviour — as the actual commodity, traded in "behavioural futures markets." Nick Couldry and Ulises Mejias's The Costs of Connection (Stanford University Press, 2019) advances the data colonialism thesis: that "the historic appropriation of land, bodies, and natural resources is mirrored today in this new era of pervasive datafication," in which apps and platforms "capture and translate our lives into data… and [sell it] back to us," framed explicitly as "designs for controlling our lives."
This strand broadly agrees that the techniques are intentional and extractive. It disagrees on register — "addiction," "persuasion," "surveillance," "colonialism" carry different causal and moral weight — and it is largely theoretical rather than effect-measuring. Which is precisely why the empirical strands below became the battleground.
Four questions remain genuinely open across the literature:
1. Effect-size resolution. Can the Orben/Przybylski–Haidt dispute be settled by methodological convergence — agreed social-media-specific exposure measures, objective rather than self-reported screen time, sex- and puberty-stratified analysis — or is it irreducibly a disagreement about which magnitude counts as policy-relevant?
2. Causation versus correlation. Where trends and correlations are agreed, what experimental or quasi-experimental designs could establish that engagement-ranked feeds cause population-scale affective change rather than merely correlate with it?
3. Mechanism opacity in labour markets. Is Dubal's individualised-wage contention empirically verifiable when the pay formulas are proprietary and firms deny per-worker personalisation — and what disclosure or audit regime would resolve it?
4. Does the thesis survive the state strand? Where harms present as failures and false positives rather than as successful operator-intended change, does "it reliably works at population scale" hold — or does it need narrowing to "the apparatus is built and, in controlled cases, demonstrably effective"?
1. Does the critique side have a home yet? The named institutes are real, but is "critical data studies" — or "technology & society," or "trust & safety" — becoming a degree-granting discipline with a shared canon, or will it stay a confederation of reading lists?
2. Is the seam closing? FAccT and crossings like the CITP dark-patterns paper suggest some traffic between build and critique. Is there measurable co-publication and cross-attendance, or do the two sides still mostly cite past each other?
3. Who teaches the ethics to the builders? If the loop-building curriculum lives in business schools and paid practitioner courses, where in that pipeline — if anywhere — is the critique taught to the people who will actually ship the systems?