Chapter 1. Equal on Paper

1.1Records and method

The records come from the public catalog of Cosmic Press, which the studio serves as a manifest and a set of numbered detail files. The manifest names 24 topics and 768 entries and gives the catalog's source as "AI-generated examples"; each of the 768 entries also marks itself AI-generated. An entry has a question in the voice of a person, a compressed general form of it, a context, a need, an assumption, a line on what the compression keeps (preserved), a line on what it loses (lost), a counterquestion, a defect and six scale notes. A record has no options and no judgment, and it never answers its question.

The studio files 34 entries under Justice. They are spread unevenly over the catalog's stretches, which differ in manner: 12 lie in the catalog's first 240 entries, 21 in the long middle stretch from q241 to q744, and 1 in the closing stretch. The selection rule is one sentence: every third Justice entry in the catalog's order, counted from the first. It takes 12 entries, q06, q133, q136, q139, q246, q318, q390, q462, q534, q606, q678 and q750, of which 4 lie in the first stretch, 7 in the middle and 1 in the closing stretch. The rule was chosen so that the sample would cross the whole catalog without choosing records by their content, and so that no record would be one that an earlier Cosmic Press study read: none of the twelve stands in those studies' sets. The record identifiers are plain serial numbers, and no field of the twelve names a model or a family of models.

The stretches show on the page. In the first stretch the records are terse: the shortest question of the twelve, "Why should I follow rules I did not choose?", runs to 25 Korean characters, and the scale notes are mostly brief statements. In the middle stretch the questions lengthen, to 87 characters at most, fill with sensory and emotional detail, and all six scale notes of 7 of the twelve records are questions. Three records carry labelled scale notes: two in the middle stretch with the labels Individual, Relationship, Society, Other life, Planet and Far future, and the one closing-stretch record with Individual, Society, Humanity, Life, Earth and Universe. The study does not know how these manners were made and claims nothing about it. It treats them as manners that can be told apart, and it asks whether the way a Justice record frames unfairness holds across them.

Cosmic Press publishes its records in Korean. The English in this paper is a translation made once by the editorial session of this study, sentence by sentence, with nothing added; each case's English seed names the Korean original by its hash, and the appendix gives the English of the fields the paper reads. Every block quote is that English, word for word. Every count is made on the Korean by a plain pattern named beside it in the study's numbers, so that the translation does not decide it. The patterns are crude word tests, and Section 3.3 says where they can mislead.

All counts come from a script (scripts/count.mjs) run on the kept records and a snapshot of the whole catalog, both fetched on the same day, with all 12 records matching the snapshot word for word. The topic label and the vocabulary agree: 10 of the twelve use a word of fairness, justice, equality or legitimacy in the question or its compressed form, as do 24 of the 34 Justice entries, against 64 of the 768 entries in the whole catalog. Of the twelve, 7 say in their line on what the compression keeps that it keeps a difference, a gap or a conflict; 5 fault their framing for missing a structure, a policy, a regulation or a rule; 5 bring a law, an institution or an imposed duty into the scale notes; 2 counterquestions ask who.

The script also takes the series' frame measure: for each field, the share of its five-word runs that recur in at least 5% of the records that fill it. Every field measures 0% in Korean, in the whole catalog and in the Justice topic, and every field of the twelve in English measures 0%. The records do not repeat sentences, and no field is set aside as frame. What the form guarantees is a different matter. Every record has a counterquestion and a defect and a line on what was lost; that a record criticises itself is the form's doing and is not a finding. What this paper reports is what the records put in those places.

All 12 contexts call their situation hypothetical. The records mention museums, workshops, phones, mines, housing laws, councils and vampire bats. None of that is information about the world, and nothing in the records is legal, financial, housing, medical or safety guidance. The paper treats each sentence only as part of a hypothetical record's framing, and it takes no side in any dispute a record describes.

1.2An equality on paper and the opportunity in use

The records place unfairness in a gap. Their questions name a sameness that holds on paper and ask whether it holds in use. The plainest case is an AI-generated record about a hypothetical day worker who wants to visit a museum:

The record states the premise the question doubts in one flat sentence,

and says what its compressed form keeps:

The museum is open to all by price and closed to some by the clock. The record names the two sides of that gap in nearly the terms Scanlon, Substantive Opportunity (2017), uses for selection processes: procedural fairness is not enough, and what matters is whether a person can in fact reach the position. Scanlon writes of positions of advantage and of developing the abilities to qualify for them; the record writes of an afternoon at a museum. The fit is loose, and this study's reading takes from Scanlon only the shape of the argument, that an open door is not yet an opportunity.

A middle-stretch record, also AI-generated, sets the same gap in a hypothetical residents' workshop where beginners lose the popular tools to regular users while they learn the booking system. Its premise and its kept difference are almost a paraphrase of the museum's:

A third, an AI-generated record of a hypothetical work dinner at a barbecue restaurant where a vegetarian member is asked for an equal share, names the gap in the vocabulary of the second debate itself:

Free entry, one screen, an even split: in each record the sameness is real and the record does not deny it. What it denies is that the sameness settles the question. In this the records stand where Fredman, Substantive Equality Revisited (2014), begins, with the limits of formal equality taken as well recognised. Fredman's dimensions help sort what each record puts on the far side of the gap. The museum and the workshop speak of access and disadvantage, a working day and a learning curve that turn an open door into a closed one. The dinner speaks of accommodating difference, a diet the shared menu does not allow for.

Other records of the sample put the gap between what is seen and what is done, and here the first debate comes in. In an early record, a hypothetical amateur theatre company has finished its first performance:

The question is about recognition: the friend's work is invisible, and the remedy it implies is to be seen. The record's critique of its own compressed form, which asks how invisible contributions are recognised, refuses that remedy as sufficient:

For one hypothetical theatre company, the record voices the worry that Fraser (2007) raises about a turn in theory as a whole. In this study's reading the record takes Fraser's two-dimensional view in miniature. It does not drop recognition, since its scale notes still distinguish the joy of having one's name called, but it insists that the work must also be shared differently. Alcoff's objection to Fraser, which concerns identity politics, finds little to hold here, since the record names no group; the friend is a person with a broom and a phone, not a member of a class or an identity.

An early record about a hypothetical residents' meeting, where quick speakers carry the day, makes the same move along Fredman's dimension of voice and participation. Its premise is that silence is consent ("If one does not object at once, one has agreed."), and its critique warns that the obvious cure is too narrow:

The meeting is formally open to all speakers; the record says that slowing it down would leave a deeper inequality of standing in place.

Not every record has a sameness at its centre. The early record on rules one did not choose asks about consent and legitimacy, the record on a borrowed textbook about reciprocity between friends, the record on a fine for a missed cleaning day about how to sanction a broken rule. Even these hold a version of the gap: a rule binds all equally whether or not each agreed to it, a friendship is equal in name while one side keeps the other's book, a fine and a day's labour are treated as equivalent. But the gap is clearest, and named in the records' own words of formal and actual, in the access records of the first and middle stretches. This section adds to the debate over formal and substantive equality a description of a generated form that builds the distinction into its questions: the records do not argue that formal sameness is insufficient; they take it as the starting point of the question.

1.3Who carries the remedy, and the level the question skipped

The records' critique of their framing turns from the claim to the cost of meeting it. The museum record, having kept the gap between formal openness and actual chance, goes on to fault its own framing on behalf of people the question did not mention:

and its counterquestion asks the question the claimant's question did not:

Opening late, the remedy the counterquestion has in view, is somebody's labour. The record does not say whose, or whether the cost is worth bearing. It says only that the question of access, put by itself, hides the question of burden. The theatre record does the same in its counterquestion:

Here the burden falls back on the claimant: recognition without redistribution leaves the friend with the broom. The workshop record turns to the other side of the queue. Its line on what was lost names the nervousness of a beginner's first grip on a tool and, beside it, "the plans of the skilled people waiting for a booking"; its counterquestion asks

The phrase "new unfairness" carries the point. A remedy for one unfairness is expected to make another, and the record asks who would decide how to trade them. It leaves the asking there.

In the middle stretch the counterquestion often speaks for the party that would pay, and in a sharper voice than the claimant's. In the AI-generated work-dinner record, it is the meat-eating majority:

In the textbook record, the critique protects the borrower, warning of "over-interpreting the other's unavoidable circumstances or forgetfulness as malicious exploitation". In the AI-generated record of a hypothetical tenant in a windowless room, the line on what was lost names a cost that would fall on people poorer than the claimant:

Read through Fraser's first debate, these records keep moving a question of recognition or rights toward a question of distribution: who works the late shift, who cleans next time, whose rent rises. They do not choose between the claimant and the party who would pay. This study reads that as the form at work, not as a stance: the record has a slot for the cost of the remedy and fills it, and it has no slot for deciding.

The critique does a second thing, and it depends on the level at which the question was asked. Where the question takes a particular grievance that has no single author, the critique points to structure and policy. The windowless-room record faults its framing for its reliance on rules alone:

The right-to-repair record, about a hypothetical owner locked out of a phone after replacing its battery, lists in its line on what was lost the arguments the question set aside, "consumers' safety and of protecting intellectual property", and faults itself for not dealing with the industrial ripple effects of the legislative debate. The clearest case is the middle-stretch record that sets a consumer's hand on a smooth screen against the hand of a hypothetical assembly worker losing eyesight and joints:

This sentence sits squarely in the third debate. In the terms of Young (2006), the critique leaves a frame of guilt, close to the liability model that looks for perpetrators, for a frame of structure and regulation, closer to the social connection model. Its scale notes then carry the question outward rung by rung:

In this study's reading the three rungs move from awareness to shared action to institutions, in the order of Young's forward-looking, shared responsibility. But they move only as questions. Measured against Wolff, Structural Harm, Structural Injustice, Structural Repair (2024), the ladder names candidates for the agent a repair would need, the self, a solidarity, governments, and commits to none of them. Browne, The Untraceability of Structural Injustice (2024), asks how anyone can be responsible when responsibility cannot be traced, and proposes looking at transitions between structural and fault-based injustice. The record's critique makes one such transition in reverse: it takes a question that started as an accusation between two hands and declares it structural. It does not trace the structure back to anyone.

Where the question takes a general principle, the critique points the other way, back to the particular. The earliest record of the sample asks in general why one should follow unchosen rules, and its critique warns of the general:

What it says was lost is the context: which rule falls on whom, and with what disadvantage. The middle-stretch record on booking order for a shared drill compresses its question into one about fair standards for shared resources and then notes what that cost:

The closing-stretch record, on whether to fine a neighbour who missed the communal cleaning or give the neighbour extra work, does both. Its line on what was lost brings back the particular person:

and its counterquestion names a burden that would fall on others if the fine became the custom: "the sense of deprivation of the people who clean in person". Its labelled scale notes, statements rather than questions, end at the level of the Universe with a sentence that keeps the word of fairness and changes its object:

Fairness, at this rung, describes how energy is put into a cleaning, not how a burden is shared among neighbours. The sentence names no party. The record's critique has brought the absent household and the people who clean into view; its top rung lets them go again.

In this study's reading, then, the critique names the level the question skipped. A grievance between persons is sent up to structure and policy; a general principle is sent down to the case and the person it hides. Both moves bring someone into view who carries a cost: the operators, the friend, the regular users, the poorest tenants, the absent household. Neither assigns that cost.

1.4What this chapter adds

The chapter adds two descriptions to the debates the introduction set out. To the debate over formal and substantive equality it adds a generated form that starts from the gap rather than arguing for it: the records' questions name a sameness that holds on paper and ask whether it holds in use. To the debates over redistribution and recognition and over responsibility for structural injustice it adds an account of where the form's self-critique goes. It turns recognition and access claims toward the distribution of the burden of meeting them, and it turns each question toward the level it left out, structure for the particular and the particular for the general. In every record it stops before saying who should carry the burden, and that stop is part of what the chapter describes.