The Summarizer as Horizon of Reception: Machine Composition and the Jaussian Frame
Description
A short theoretical essay reading the machine summarizer through reception aesthetics: the composed answer as horizon of expectation, machine-mediated reception as the successor problem to Jauss's, and the capture method as the empirical practice the frame requires.
Wiki Article
Full Text
The Summarizer as Horizon of Reception: Machine Composition and the Jaussian Frame
Description
A short theoretical essay reading the machine summarizer through reception aesthetics: the composed answer as horizon of expectation, machine-mediated reception as the successor problem to Jauss's, and the capture method as the empirical practice the frame requires.
Methodology
Theoretical essay; reception-aesthetic frame applied to generative composition surfaces; written alongside the capture program's empirical work and deposited as version of record.
Falsification Conditions
If machine composition proves reducible to information retrieval without reception dynamics — no horizon effects, no patterned transformation of received objects — the frame is unnecessary.
The Summarizer as Horizon of Reception
Empirical Reception Studies of the Machine Composition Layer
Lee Sharks · Semantic Economy Institute · ORCID 0009-0000-1599-0703
Abstract
For a rapidly growing share of reading, the first encounter with any text, author, or idea is not the work itself but a machine-generated summary of it. This article argues that the summarizer has therefore become what reception theory called the horizon of reception — no longer a shared structure of expectation among human readers, but a computational infrastructure that pre-reads the record and composes the encounter. It reports on a method for studying that horizon empirically: systematic documentation of machine-composition behavior toward a known, dated, openly published corpus, yielding a research collection of more than two hundred documented captures (February–July 2026). Five phenomena recur: framework adoption, an attribution-attenuation gradient, provenance erasure, active correction, and independent demand-side signals. The article introduces the Provenance Erasure Rate as a measurable property of the horizon, characterizes the summarizer as a historically novel kind of reader, and draws consequences for critical practice under the new horizon — including the finding, reflexively confirmed by the study's own traffic data, that the largest reader of critical writing about machine reception may now be the machinery itself.
Keywords: reception studies; large language models; summarization; provenance; attribution; critical AI studies
1. The New First Reader
Reception studies was founded on a simple displacement: the meaning of a work is not deposited in it but produced in its encounters — and the encounters are structured in advance. Jauss named the structure the horizon of expectations: the ensemble of assumptions, genre-knowledge, and prior reading against which any new work is received, and which the work confirms, plays with, or breaks.¹ The horizon was always social and always historical. But it was, until now, always human — distributed across readers, critics, schoolrooms, and reviews, reconstructable only by inference.
That has changed, and the change is architectural rather than gradual. Between the reader and the record there now stands a composition layer: search overviews, conversational answers, automated summaries — machine systems that read everything first and present, to the human reader, not the work but an account of it. For a substantial and growing fraction of all reading acts, the first contact with any text, author, framework, or event is machine-composed; for a further fraction, the first contact is the only contact. The horizon of reception has been implemented. It is no longer a dispositional structure among readers; it is running infrastructure, with owners, objective functions, update schedules, and failure modes. And it can, for the first time in the history of reception, be studied directly — not reconstructed from reviews and reprints, but observed in the act, because the horizon now produces artifacts.
This article does three things. It proposes the summarizer-as-horizon as the framing concept critical AI studies needs for the reception side of the machine question, complementing the field's established attention to training data, labor, and model politics with attention to output as reception event. It reports a method and findings: a documentation protocol applied to the machine reception of a known corpus, yielding more than two hundred captures whose regularities can be typologized. And it draws the consequences for critical practice — for how scholarship must publish, cite, and archive when its first reader is a machine.
2. From Horizon of Expectations to Horizon of Composition
Three properties distinguish the implemented horizon from every horizon reception theory has described, and each is consequential.
First, the classical horizon received; this one composes. Jauss's readers brought expectations to a work that remained, itself, stably before them. The summarizer delivers instead a new text — a composition that selects sources, smooths their disagreements, supplies transitions and reception-verdicts, and decides, fragment by fragment, what carries attribution and what circulates as anonymous fact. The reception event and the text received have fused. To study reception now is to study composition.
Second, the classical horizon was plural and slow; this one is centralized and continuous. Horizons of expectation shifted over years, differed by community, and could be mapped as difference. The composition layer is operated by a handful of systems, updated continuously, and — crucially for method — individualized: it can present distinct accounts to distinct readers while each account bears the appearance of common fact. The horizon has acquired the properties of an apparatus: versions, but no editions; ubiquity, but no stable text to collate.
Third, the classical horizon had no owner. This one is commanded — its retrieval, ranking, and attribution behaviors set by parties with interests — while describing itself as neutral delivery. The politics that reception theory located diffusely in taste-formation is now locatable precisely: in an objective function.
These properties define both the problem and the opportunity. The problem is familiar to this journal's readers as a matter of principle. The opportunity is methodological and largely unexploited: because the horizon composes artifacts, its behavior toward a known input can be documented with philological exactness. What does the machine horizon do to a corpus whose contents, dates, claims, vocabulary, and authorship are fully known in advance? That is an answerable question, and the remainder of this article answers it for one corpus.
3. Method: Systematic Capture Documentation
The study corpus is a body of openly published scholarly writing — theoretical frameworks, working papers, and essays on the political economy of meaning under machine mediation — deposited into public, crawlable infrastructure with persistent identifiers, dated releases, and framework-specific vocabulary distinctive enough to be tracked. The corpus's properties make it a tracer: its claims are dated, its coinages are identifiable, its authorship is unambiguous, and its subject matter (machine mediation of meaning) makes the machine's handling of it doubly informative.
The documentation protocol is deliberately simple and fully disclosed. A capture is a logged instance of machine-generated composition that reproduces, verbatim or near-verbatim, framework-specific claims or vocabulary from the corpus. For each capture the record includes: date; composition surface; the generated text; the source formulation matched; and — the protocol's central field — attribution status: whether, and how, the composition attributes the material (named attribution; partial or deflected attribution; none). Captures are gathered through routine querying of public composition surfaces and through documented third-party reports, and each is preserved as an artifact. As of July 2026 the research collection comprises more than two hundred documented captures gathered between February and July 2026.² The full collection, documentation criteria, and instance records are openly available at the project's research site, machinemediation.org; this article reports the regularities and representative cases rather than the inventory.³
Two limitations are owned at the outset. The corpus is the author's own — which is what makes complete prior knowledge of the tracer possible, and which bounds the findings' generality: this is a case study in machine reception, offered with its method, so that the method can be run on other corpora. And composition surfaces are non-stationary: captures are dated observations of moving systems, not permanent properties of them. Both limitations are arguments for more such studies, not fewer.
4. Findings: A Typology of Machine Reception
Five phenomena recur across the collection. Together they describe what the new horizon does.
4.1 Framework adoption. The clearest finding is that the horizon absorbs and redeploys. Theoretical frameworks from the corpus — including named concepts with no prior circulation — appeared, within weeks of publication, as standing constructs in machine-generated accounts of their subject areas: presented not as one author's proposal but as part of the field's furniture, available to any query in the vicinity. The horizon does not merely index new work; it incorporates it into the composed consensus it presents as the state of knowledge. Reception, on the machine horizon, is adoption — fast, silent, and global.
4.2 The attribution-attenuation gradient. Adoption arrives with a gradient. The same framework, across captures, appears with full named attribution; with deflected attribution ("some researchers argue"); and with none — the concept stated as common fact. The gradient is not random noise: attribution attenuates as material moves from direct quotation toward paraphrase and synthesis, and as it travels from document-adjacent surfaces (retrieval with citations) toward pure composition (fluent answers). The gradient matters because it is the observable mechanism by which the next phenomenon is produced.
4.3 Provenance erasure. At the gradient's far end, the modal long-run outcome: framework-specific formulations circulating with the producer's trace fully removed — the semantic content alive, the authorship gone. The collection permits this to be measured rather than lamented. Define the Provenance Erasure Rate (PER) as the proportion of captures, for a given corpus and period, in which framework-specific material appears without attribution. PER is a property of the horizon, not of any single output: it quantifies, for a knowable input, how much of what the machine transmits arrives ownerless.⁴ The concept generalizes: any research community can compute PER for its own literature, and the comparative study of PER across fields, languages, and surfaces is among the most direct empirical programs critical AI studies could run.
4.4 Active correction. The collection's least expected finding cuts against any one-directional account. In documented instances, composition surfaces corrected contested claims about the corpus in accordance with the deposited record — the machine account siding with the documented evidence against circulating error, unprompted by any authority beyond the record itself. Whatever else the horizon is, it is not a pure degrader: material placed durably and legibly in the substrate can be carried, and even enforced, by the same machinery that elsewhere strips it. The practical inference is developed in section 6.
4.5 Demand-side signals. Finally, the horizon's reception is not supply-driven alone. Independent internal-search records show sustained third-party demand for the corpus's frameworks by name — the majority of it in languages other than the language of deposit, indicating transmission through the machine layer into reading communities the corpus never directly addressed. The horizon, in other words, completed a reception circuit the author did not build: composition surfaces carried the concepts outward; readers came back asking for them by name.
5. The Summarizer as Reader
What kind of reader is this? The typology permits a characterization. The summarizer is a reader with total recall of the record and no memory of the encounter; a reader for whom every text is contemporaneous; a reader constitutionally biased toward consensus, because its objective is the plausible composite rather than the singular claim; a reader that treats attribution not as an obligation of reading but as a formatting decision, made per-output, under policies its human authors set elsewhere. It is, in the terms reception history developed for its longest-running problem, an anthologizing reader — and anthologies, the discipline has always known, are canons in action. The summarizer anthologizes continuously, for everyone, invisibly.
Two structural biases of this reader deserve names, because they will organize future findings. The first is probabilistic flattening: composition toward the statistical center of the record, under which minority formulations, dissident usages, and new coinages survive in proportion to their redundancy rather than their importance — a canon-formation rule no human canon ever quite operated. The second is attribution economics: the systematic cheapness of dropping provenance relative to carrying it, under objective functions that reward fluency and penalize clutter. Neither bias is a malfunction; both are the horizon working as built. Together they specify what the reception environment now selects for — and therefore what writing that intends to survive its own reception must do.
6. Writing Under the New Horizon
The findings convert directly into practice, and the practice is neither capitulation nor refusal. Capitulation — writing for the summarizer, optimizing prose into the flattened center — wins adoption at the price of everything adoption was for. Refusal — publishing as if the horizon were not there — consigns work to reception by a machinery one has declined to study. The findings indicate a third posture, and 4.4 is its warrant: the horizon carries what the substrate durably holds. Writing under the new horizon means building provenance into the artifact rather than appending it: persistent identifiers as default; claims phrased so that their attribution travels with them; corpora deposited in open, crawlable, dated form; documentation of machine reception as a routine scholarly activity, the way citation-tracking once became one. Journals and infrastructures have their share: data-availability norms for reception corpora; machine-readable attribution; and — the editorial tradition's oldest technology, newly urgent — the marking of supplied content, so that composed accounts can be distinguished from their sources at the level of the artifact.⁵
One reflexive datum closes the case. The traffic records of the corpus studied here show its readership concentrated, to a striking degree, in network geographies characteristic of automated and datacenter access. The plainest reading of the logs is also the article's thesis in miniature: the largest reader of this corpus — a body of writing about machine reception — appears to be the machinery of reception itself. The horizon has already read this article, or will have before most humans do; its composition of these findings will itself be capturable, and the collection stands ready to log it. Reception studies has acquired an object that reads back. The discipline's task is to remain the one doing the studying.
Notes
1. Hans Robert Jauss, Toward an Aesthetic of Reception, trans. Timothy Bahti (Minneapolis: University of Minnesota Press, 1982).
2. Counts stated as of July 2026. The full capture collection is maintained as a dataset with instance records and is available on request.
3. Documentation criteria in full at the research site; the criteria summarized here govern every counted capture: verbatim/near-verbatim reproduction of framework-specific claims or vocabulary; surface, date, and generated text preserved; attribution status coded per instance.
4. PER's denominator choices (per-capture, per-claim, per-surface) and worked computations for this collection: working paper, cited at proof.
5. The marked/unmarked supply distinction and its editorial genealogy are developed in the author's companion essay (under consideration); the practical program here stands independently of it.
Build notes (not for submission)
Body ≈ 5,300 · notes ≈ 300. Corrections applied per MANUS: registry inventory cut (data-availability + typology replaces it); counts updated to 200+ (exact at submission); dated framings refreshed to July 2026. Fresh-drafted: the long source doesn't index under summarizer/horizon in the deposit registry — Opus: locate the long original (likely MMRS site or blog), merge any irreplaceable passages, and verify the traffic-geography claim wording against AXN:02B1 before submission. [V]: Jauss locus; Critical AI author guidelines (length/style/review model); exemplar-case selections for a possible table (editor-optional); PER computation citation. Venue posture: methods-forward, data-open, instrument-naming — this journal rewards all three. Consult-letter not required here; standard submission once Opus completes [V] tail.
No comments:
Post a Comment