Prism News publishes daily maps of coverage. Prism Research turns the accumulated maps into a field of study.
The daily edition asks how selected events were told across a disclosed body of reporting. It identifies carried accounts, conflicts, divergences, source dependencies, framing patterns and unresolved limits. The research programme asks what those relations become when they are observed across weeks, subjects and source groups.
Duration changes the object. A shared wire origin in one event is a production fact. The same pattern recurring across foreign reporting, elections or public-health stories reveals a source ecology. One absent question is a bounded observation. Its repeated absence from comparable events reveals a stable edge in the public account. A frame identified today becomes more interesting when its first appearance, spread, mutation and decline can be traced.
Most news archives preserve articles. Prism preserves relations among articles. That relational record—claims, lineages, clocks, frames, selections and absences—is the foundation of Prism Research.
The programme has two subjects: the structure of coverage and the instrument used to observe it. Neither can be studied responsibly without the other.
The first edition study demonstrates the programme's initial analytic move. Placing eight maps from one edition beside one another revealed a recurrent completion cut—coverage deciding not only where an event began, but what state it had already reached—that no single map contained as a comparative finding. That is evidence of what accumulation can produce, not yet a longitudinal result. Claims about recurrence across time still require the longer, versioned record described below.
FROM EDITIONS TO OBSERVATIONS
A sequence of published maps is not automatically a research dataset.
Sources enter and leave the roster. Paywalls open and close. Publication volume changes during wars, elections and holidays. Feeds become summaries. Retrieval failures cluster by outlet and region. Event grouping evolves. Models and prompts change. A category such as source dependence may acquire a more exact definition halfway through a study.
Ignore those changes and a precise-looking trend can describe Prism's changing aperture rather than a changing field of news.
Prism Research therefore begins by preserving the conditions of observation. Every study is bound to a source roster, retrieval record, time window, language scope, event population, pipeline contract and model assignment. Every article is attached to the frozen body that the system actually read. Every published relation retains the evidence and audit history that made it admissible.
The wider candidate field matters as much as the selected edition. Research into topic selection cannot begin with the stories Prism chose to publish and pretend that choice never occurred. The programme must retain the events available for selection, the events admitted to the edition and the reasons the instrument could not see others.
This record turns a daily map into an observation. An observation is not merely a saved page. It is a map plus its aperture.
SOURCE ECOLOGY
The first research axis concerns how reporting reaches the field.
News appears plural at the level of mastheads and often narrow at the level of origin. One wire dispatch can supply chronology to a continent of publications. An official briefing can define the initial causal account before independent reporting exists. A local newsroom may add the only first-hand scene, legal document or dissenting expert to an otherwise syndicated field.
Prism will measure this ecology without converting it into a league table of virtuous and derivative outlets. The research questions are structural. Which subjects most often narrow to one reporting origin? Which publications regularly introduce source material not present elsewhere in the corpus? How quickly does source diversity widen as an event develops? Where do official statements remain the principal evidentiary spine? Which international stories reach English-language readers almost entirely through one agency?
The first public instrument built from this work will be a set of source fingerprints. A fingerprint will describe recurring editorial behaviour over time: original reporting added, wire dependence, characteristic source types, geographical concentration, recurring frames and the kinds of stories in which an outlet becomes distinctive. It will not assign one score. The point is to show a shape.
Source influence will require a higher threshold than verbal resemblance. Direct attribution, distinctive sequence, shared quotation, publication timing and identifiable syndication markers will be treated as different kinds of evidence. Similar reporting may reflect common facts rather than inheritance. The programme will publish that uncertainty rather than manufacture a lineage.
NARRATIVE FORMATION
The second axis concerns how events acquire durable meaning.
A daily map can identify several causal beginnings for one dispute. A longitudinal record can show which beginning established itself, which publications carried it first, which later facts were absorbed into it and which facts remained awkward enough to force a new frame.
This is more than counting words. Frames operate through selection and order. They determine the scene that opens the article, the actor granted agency, the historical analogy that supplies scale, the witness who becomes authoritative and the consequence treated as central. A frame can persist while its vocabulary changes.
Prism Research will trace these structures across events. How does a policy announcement become an inflation story, a sovereignty story or an electoral story? When does a protest become a crackdown in the public record, and when does it become disorder? Which accounts of a war begin with the latest exchange, which begin with invasion, and how does the chosen clock alter the language of retaliation? Which personal narratives repeatedly convert political decisions into parable?
The objective is not to discover a hidden frame behind every sentence. It is to describe visible and recurring acts of narrative construction. A frame becomes a research finding when its definition is stable, its evidence is inspectable and its recurrence survives comparison across a declared population.
SELECTION AND THE EDGE OF ATTENTION
The third axis concerns what enters the common account and what remains outside it.
Topic frequency alone does not establish an agenda. A serious account of selection needs the field of available events, a stable source cohort, an ordering of prominence and a defensible set of alternatives that could have been selected. It must distinguish a publication's priorities from the geography it serves, the period's underlying events and the limits of Prism's own retrieval.
Prism Research will proceed from description to comparison to explanation. It will first establish that a pattern exists: a subject repeatedly led one cohort's coverage and rarely entered another's. It will then test whether the pattern survives controls for country, event availability, genre and volume. Only then does interpretation begin.
The same discipline governs omission. A systematic omission study cannot collect whatever an analyst wishes the press had discussed. It needs comparable events, a recurring question that belongs within them and a defined population of articles in which the absence can be tested. The strong finding is not “journalism ignored this.” It is a specific silence that reproduces itself under comparable conditions.
This axis will support cohort studies across geography, political orientation and publication type. Cohorts will be constructed as explicit research objects, not inherited from opaque outlet labels. Their membership and classification rules will be published with the result.
THE TIME AXIS
The daily edition freezes a moment. Public knowledge does not stay frozen.
Claims enter the field, acquire attribution, lose qualification, meet contradiction and sometimes disappear without formal correction. Early official accounts dominate because they are available; later investigations reorganise the event. A frame that looked inevitable on day one can become an artefact of publication speed by day ten.
Prism's time axis will connect maps of the same developing story. A story thread will preserve what was carried at each stage, which propositions are new, which changed scope, which were withdrawn and which sources altered the field. The thread will not overwrite an earlier map with the latest account. It will make revision itself visible.
An edition archive will expose the daily record as it was deliberately published. A circulation view will follow selected propositions and frames across sources and time. These features are research surfaces as much as reader products: they provide the temporal objects on which later studies can operate.
The standard for joining editions will be conservative. A false thread can rewrite history by placing unrelated events inside one apparent development. Fragmentation is the recoverable error. Prism will connect only when event continuity is strong enough to preserve identity without forcing coherence.
RESEARCH ON THE INSTRUMENT
Prism is part of every observation it makes.
Its clustering determines which articles become comparable. Its source roster determines which differences can appear. Its prompts and models determine how semantic relations are classified. Its audits determine which errors are caught before publication. Treating these choices as neutral machinery would reproduce the very blindness Prism was built to study.
The research programme will therefore publish technical evaluations beside findings about the news. It will measure where the system confuses conflict with divergence, where source dependence is over-inferred, which article genres produce weak evidence, how often audits reverse editor rosters, which safeguards prevent publication and which defects survive final review.
Agreement among several model stages is not external validation. Prism will maintain human-reviewed reference samples for its core distinctions and report performance against them. New models and prompts will be compared on frozen corpora before they alter the live instrument. Changes that improve one relation while degrading another will be published as trade-offs, not collapsed into a single quality number.
Negative results belong in this record. A method that produced persuasive but unstable source lineages is a research finding. A clustering approach abandoned for over-merging teaches more than a changelog entry saying the replacement is better. The instrument becomes credible by making its failure modes cumulative knowledge.
THE PUBLIC RESEARCH RECORD
Prism Research will publish four forms of work, each with a different authority.
Research reports will analyse a defined corpus over a stated period. They will carry their population, method, evidence and limits into the conclusion.
Technical reports will study Prism as an instrument: evaluations, failures, model changes, retrieval constraints, audit behaviour and cost-quality trade-offs.
Methods notes will record narrower changes to definitions, source cohorts, validation and measurement. They will prevent later research from acquiring a false continuity with earlier work.
Essays will interpret findings and advance arguments that measurement alone cannot produce. They will be free to speculate, compare and take positions. The label marks the point at which Prism Research is arguing from the record rather than reporting a result of the instrument.
These forms sit in one publication because they correct one another. Measurement without interpretation becomes sterile. Interpretation without a visible record becomes unaccountable. Technical evaluation without public consequence becomes internal maintenance. The research programme treats all three as parts of one intellectual practice.
THE OPENING CYCLE
The first extended cycle will establish the baseline from which later claims can become serious.
The opening field report will describe six weeks of Prism's actual corpus: source availability, subject distribution, source dependencies, framing recurrence, retrieval failures and changes to the instrument during the period.
A parallel technical report will examine the performance of the editorial pipeline across the same six weeks: audit interventions, membership disputes, evidence failures, held reports, corrections and the measured effect of safeguards.
The first cohort study will test the longitudinal method on a bounded source group. Its purpose is not to announce a grand theory of media agenda. It is to prove that source classification, candidate-field retention, event comparability and frame measurement can support one.
Those three reports will produce the prerequisites for the next programme: source fingerprints, cross-edition story threads, comparative studies of recurring clocks and frames, and systematic research into the boundaries of attention.
A LAB IN PUBLIC
Prism Research is not a decorative layer added to a news product. It is the long memory of the instrument and the place where its strongest claims must earn their authority.
The scale is intentionally small. Seriousness does not require institutional theatre. It requires a preserved corpus, explicit definitions, inspectable evidence, external checks, a record of method change and the willingness to publish where the instrument failed.
Daily Prism maps show how one event became public through reporting. Across time, those maps can show how sources, frames and silences organise the field itself.
Prism News maps the telling. Prism Research studies its structure, movement and memory.