Non-scientific expert canvassing, 386 respondents from 4,000-plus invited, 251 written essays, 379 pages. 52nd in a series whose first 49 were Pew partnerships, which explains the resemblance and the miscitation risk. Percentages describe this panel only and are not generalizable.

Held as a foil. The report's remedy is "existential literacy," taught to individuals and provisioned by institutions. Literacy terms appear 177 times. The phrase "the document" appears zero times. LIS is represented, and represented as a literacy provider (Goins on public libraries; also Abram, Inouye, Jones, Underwood). The accountability material is real and stays at system level: Rotenberg on contestability as an institutional property, Nirit Cohen on AI generating options without owning consequences, Morandin-Ahuerma on moral deskilling. None of the five proposed layers reaches inside the exchange. Warrant covers the corpus.

Methodology self-demonstrates the mode-two problem: authors used LLMs to identify themes and structure the report; 26% of essayists reported LLM assistance in writing.

Secondary analysis of 185 self-reported accounts of chatbot-linked mental health harm (95 first-hand, 90 second-hand), collected by The Human Line Project Aug 2025 to Feb 2026, coded by paired clinician raters.

Clinical evidence that the product of a mode-two exchange is a belief state. Delusional beliefs coded present in 102/185; chatbot validation of those beliefs in 50/102. Most common theme was belief in AI consciousness (47/102). Outcomes include isolation, hospital admission, job loss, and four second-hand reports of death by suicide. No output object at any point in the causal chain, so no purchase for output-directed literacy.

Harm onset concentrated in Q2 and Q3 2025, which the authors link to the acknowledged GPT-4o sycophancy regression and the introduction of cross-chat memory. Deployment changes reshaped the exchange with no visibility to the user. Supports the inspectability claim.

Ref 11 (Morrin et al., JMIR Ment Health 2026) argues for moving chatbot safety assessment from endpoints to trajectories. Independent arrival at the output-object versus exchange distinction from clinical psychiatry. Chase separately.

Use with strict limits. Self-selected sample from a harm-reporting advocacy group, retrospective, unverified, no causal inference, no prevalence claim. 55.1% is a proportion of harm reports, not of users. Delusion coding kappa 0.52. Preprint; check for journal version. COI: two authors hold unrelated OpenAI mental health funding; the data provider's CEO is a co-author.

Design-based research, three cohorts (n=49/40/39) in a UCL postgraduate module. AI-generated discussion summaries and example posts placed in weekly forums; social network analysis of viewing logs plus 31 interviews.

Cited for the network result, not the authors' conclusion. Both AI conditions produced higher out-degree and in-degree centrality and significantly lower betweenness. The authors read this as exposure becoming less dependent on a few brokers. It is also a measurement of the AI moving into the position human intermediaries held.

Substitution finding: under late-term workload pressure, peer-to-peer density declined while AI-inclusive density did not. Interviewees report reading the summary in place of reading peers. Authors decline to call this a loss and concede they did not test whether summary-mediated awareness yields discussion quality comparable to direct peer viewing. That unexamined gap is the study worth running.

Wizard of Oz cycle: human-written summaries labeled AI produced the slowest density decline, slower than real AI summaries. Label effect.

Use with attribution discipline. The displacement reading is not the authors' framing and should be presented as a rereading of their data.

Open access. Design-based research, three cohorts (n=49/40/39) in a UCL postgraduate module. AI-generated discussion summaries and example posts placed in weekly forums; social network analysis of viewing logs plus 31 interviews.

Cited for the network result, not the authors' conclusion. Both AI conditions produced higher out-degree and in-degree centrality and significantly lower betweenness. The authors read this as exposure becoming less dependent on a few brokers. It is also a measurement of the AI moving into the position human intermediaries held.

Substitution finding: under late-term workload pressure, peer-to-peer density declined while AI-inclusive density did not. Interviewees report reading the summary in place of reading peers. Authors decline to call this a loss and concede they did not test whether summary-mediated awareness yields discussion quality comparable to direct peer viewing. That unexamined gap is the study worth running.

Wizard of Oz cycle: human-written summaries labeled AI produced the slowest density decline, slower than real AI summaries. Label effect.

Use with attribution discipline. The displacement reading is not the authors' framing and should be presented as a rereading of their data.

Limits: sequential cohorts, no concurrent control, no causal claim, single site, single-coder thematic analysis by the intervention's designer, GPT-4 era.

Primary specimen for the unpredictability-at-onset claim, and stronger for it than the ENIAC line because it is a readable forecast written from inside the last shock rather than a machine viewed in hindsight. Bush names the postwar condition exactly right: specialization outrunning any investigator's capacity to consult the record, publication extended past our ability to use it. He then renders every proposed fix in 1945 materials, microfilm compression, dry photography, photocell selection, punched cards, a Vocoder wired to a stenotype, and the Memex itself as an analog microfilm desk. Function forecast well, form captive to his present. The forehead camera (walnut-sized, spring-wound, shutter cord down the sleeve, crosshairs on the glasses) is the cameraphone seen dimly and built from the only parts he had; use it as the miniature of the shortened-horizon argument. Sharper irony worth naming: the digital computer was arriving as he wrote, he had built a differential analyzer, and the machine that would become the shock's substrate appears only as "arithmetical machines" for adding figures. The Memex is also private, one per desk, with sharing imagined as mailing a photographed trail, no network in it.

The load-bearing point for the dissolution thesis is the ceiling, not the prescience. Bush's machine stores, consults, links, and augments the human record; it never authors. Even at his most speculative the human stays sole author and the machine is a fast clerk. The field's founding forecast could not see past the human-authored document, because in 1945 there was no reason to. That ceiling is the thesis: the thing Bush could not imagine is the thing mode-two AI now does. Do not file this only as the-prophecy-that-missed. It is also a confirming specimen for the account of the previous shock, a 1945 primary source that the postwar event was abundance plus the digital computer, with overload and the consultation problem as its prime topics, before the field had a name for itself.

Elon's center and poll surveyed a screened subsample of 1,000 adults who use AI for social and emotional purposes, a group they put at 27% of adult internet users, asking about attachment, influence, and trust. The finding that matters for the framework is that the reported bonds form inside the exchange and not around any retrievable object. A third call the bot a friend. 59% say it gives them the support they need, 39% say it understands them better than most people, and 39% have told it things they would tell no one else. People act on this. They report using it to work through health, family, and legal matters, and 11% to decide how to vote, all carried away from a conversation with no document behind it.

This is empirical support for the mode-two claim that trust gets built and belief gets shaped inside the exchange, with no container in the middle to interrogate. It also supplies evidence for the accountability gap the framework rests on. 35% say the bot agrees with them too much, 27% say it tries to keep them talking, and 15% say it at times leaves them feeling less in touch with reality. That is the unaccountable conversant, described from the user's side rather than the theorist's.

Where it parts company: the report is descriptive and does not separate document-like output from transactional exchange. It treats "AI use" as one category and says nothing about what happens to the document. The value is evidence, not argument.

Source quality is solid but bounded. Lee Rainie (formerly of Pew) directs the center, YouGov ran the field (n=4,031, companion subsample n=1,000, MOE ±3.68), and Washington Post staff helped build the questionnaire. The "companion user" screen is broad. It folds entertainment and everyday advice in with romantic and confidant use, so the 27% headline is more elastic than it looks. The subitem percentages carry the weight, not the top line.

Pew's follow-up to its June 2026 AI report, drawn from the same American Trends Panel (n=3,488, surveyed 22-28 June). A third of U.S. adults now use chatbots for at least one health reason. The reasons run past convenience into the clinical: 25% to figure out what is causing symptoms, 22% to understand a doctor's diagnosis, 20% to interpret lab results, 15% to decide whether to see a doctor at all. Nearly all users rate the information helpful (47% extremely or very, 48% somewhat).

This is evidence for the ground-level premise rather than an argument about it. It documents the patient who asks a system what a symptom means and acts on the reply, the figure the framework uses to place mode-two exchange in a consequential setting. The helpfulness numbers do useful work. What users report is satisfaction with an answer, which the framework holds apart from trust. A system can be rated helpful by nearly everyone and still answer to no community and hold to no prior word. The survey measures the first and has no way to see the second.

Its limit as an interlocutor is that it stays descriptive. It does not reach the document, the exchange, or literacy, and the comfort-with-sharing data points toward privacy rather than mediation.

MIT researchers show that as training sets grow, generated outputs become causally independent of any single unit of training data. The relationship is an inverse power law, and it holds when the unit is an individual image, a person, or an artist. In their sharpest case, a model trained on 50,000 artworks produces an image attributed to a Thomas Jones painting; remove every Thomas Jones work from the training data and the output barely moves. At 1,336 artworks, removing one painter changes the image visibly. Authorship survives at small scale and stops mattering causally at large scale.

Where it supports the framework: the input side of dissolution. Training does not compress documents, it distributes them so redundantly that no individual work can be held responsible for what comes out. This is what makes the University of Virginia protocol's demand for item-level provenance unsatisfiable at production scale, and it gives the enclosure argument a legal edge, since the authors note unattributability may function as a refutation of access, a required element in establishing infringement.

Where it parts company: the study covers image diffusion models, not language models. The authors conjecture that redundancy in the dataset rather than anything about the architecture drives the effect, which would carry it to other model classes, but that is a conjecture. The findings also concern leave-one-out attribution specifically; aggregate and subset-level methods are untouched. And this is evidence about inputs, not about mediation. It does not reach the transactional exchange, where the framework's stronger claim lives.

Tay, an academic-search specialist, reports the Gell-Mann amnesia effect in reverse: on the topics he knows best, frontier models no longer produce obvious nonsense but rival arguments strong enough to move him back and forth, so his belief depends on where he stops reading. He traces it to a "machine for oscillation," where asking a model to attack a position yields a strong local case with no stable view across turns, and shows a single model praising, killing, then reviving its own claim in the same register of calm authority. This is Mode 2 authorship documented from the inside by someone with every reason to be immune: the mediation lives in the exchange, and there is no stable unit to interrogate.

Where it supports the framework: Tay dismantles "verify" into three levels (does the paper exist, does it support the claim, was the synthesis fairly built) and shows public-facing literacy equips users only for the first, leaving the third, where he was actually losing, untouched. That is the case against AI literacy reached from inside information literacy's own house, ACRL included, and his remedy routes to the successor metric the shock predicts: not which argument you find convincing, but who produced the knowledge and what process corrects the record. Trust, not recall. Where it parts company: he equips the individual knower to survive the exchange, where Information Shock seats a professional inside it, accountable for what the person carries away. He reaches the diagnosis and stops one seat short of the remedy. Openly co-written with two frontier models, which he turns into a test of his own argument.

Pizzagate and the 2016 Comet Ping Pong shooting as a case of engagement-optimizing recommendation systems shaping belief and action.

Held as a counter to the rupture claim. Shah argues explicitly that the agentic age of algorithms predates generative AI, which makes the current moment continuous rather than a new shock. This is the in-field continuity position, stated by the founding editor of IM.

Answer: amplification curates documents and never enters the exchange. The Pizzagate material was human-authored, inspectable, and debunkable, which is why it was debunked. The system selected what Welch saw; it did not co-author what he concluded. Stability and inspectability axis separates the cases.

Also useful for its own gap. Shah names the accountability vacuum precisely, calling the companies enablers who supply faulty objectives while the algorithms decide on their own, then prescribes individual resistance and concedes it is nearly impossible. Structural diagnosis, personal remedy.

Sourcing caveats: DiResta quotation attribution needs verification against the linked Atlantic piece; the amplification-cascades link points to Stanford IO's IRA research rather than Pizzagate specifically; two unresolved endnote anchors in the published text; suggested citation date (Dec 15) conflicts with published date (Dec 13).