Chapter 1. Compressing the Cosmic Question

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 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. The scale notes always come in the same order. Read across the catalog, in this study's reading, the first concerns the person, the second the society and its institutions, the third plural traditions or regions, the fourth life beyond the human, the fifth the Earth or a planet, the sixth the cosmos or the largest frame. A record has no options and no judgment, and it never answers its question.

The selection rule is one sentence: every entry the catalog files under Cosmos among its first 256 entries (q01 to q256). The catalog holds 34 Cosmos entries in all; 13 fall in the first 256, and those 13 are the records read here. The rule has a reason beyond convenience. The first 256 entries are written in a different manner from the later ones, and the difference shows in the sixth scale note. By a plain test (whether the note contains a Korean negation or a verb of separating), 153 of the 243 entries outside Cosmos in the first 256 end their ladder with a sixth note that denies, limits or separates, against 63 of the 512 entries after them. The first 256 are therefore the part of the catalog where the top of the ladder has a settled job, and the Cosmos records within them can be compared with the rest of that part. The later 21 Cosmos entries are left out; Section 3.3 says what that costs.

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. Every block quote in the paper is that English, word for word. The paper counts in Korean where a count turns on the wording, so that the translation does not decide it.

All counts are made by a script (scripts/count.mjs) from the kept records and a snapshot of the whole catalog. Besides plain counts of the catalog and of the thirteen, it takes the frame measure of the series (spine §6): for each field, the share of its five-word runs that recur in at least 5% of the records that fill it. Every field of the catalog measures 0% in Korean; Korean joins particles to words, so the Korean measure is a floor, and a check with three-word runs, which is not the spine's measure, also gives 0% in every field. The thirteen in English also measure 0% in every field. The record set's frame is therefore not in repeated sentences. It is in its slots: every record has a compressed form, a lost line, a defect line, a counterquestion and six scale notes in a fixed order. Those slots are guaranteed by the form, and the paper does not report their presence as a finding. What it reports is what the records put in them.

One more caution belongs here. All 13 contexts call their situation hypothetical or a thought experiment, and 8 of them also say what the record is not: not the statement of a religion or a theory, not a real transmission plan, not a real discovery, not a real mission. The records mention stars, signals, probes, planets and the Moon. None of that is information about the universe, and this paper treats it only as the setting of a hypothetical question.

1.2What the compression removes

The compression takes the asker out of the question. Of the thirteen AI-generated questions, 7 are spoken in the first person, by a speaker who stands somewhere in the question. The count is made in the Korean, where a first-person word (the equivalents of "I", "me", "my" or "we") is present in 7 questions; the Korean often leaves the pronoun out, so the English renders more of them with an "I" or a "we". Of the 7 compressed forms, 1 keeps that person: the record on loneliness, whose general question still asks about "our lives". The other six move from a speaker to an office, a category or a capacity.

The clearest case is the hypothetical greeting sent into space. The question is spoken by someone who has been left out:

The compressed form keeps the topic and drops the person:

Two things change at once. "me, who has not yet agreed" becomes a "shared name", and the question of whether one may be represented without consent becomes a question of who holds an authority. The first is a complaint; the second is a question of jurisdiction. Tversky and Kahneman showed shifts of preference when one problem is framed in two ways; here, as a lens only (the records report no preferences), the record frames one question two ways and the second frame no longer has anyone in it who could object. The record does not leave this unsaid. Its defect line names exactly what the jurisdictional form crowds out:

and its lost line names what the general form cannot hold at all:

This is the pattern of the whole set. The lost line and the defect line are two different accounts. The lost line lists what is not in the compressed question: things particular to the case, often concrete. The defect line says what the compressed question does wrong because of its shape. Of the thirteen lost lines, 5 name something actual or real (in the Korean, the word for "actual" or "real" itself), and 5 name an uncertainty, a state of examination or a possibility. Of the thirteen defect lines, 5 name the act of compressing, bundling or reducing as the fault, 4 say that two things have been put into one word or one problem, and 2 say that asking only one thing leaves another out. The record set, in other words, gives two vocabularies for its own abstraction: one of things that dropped out and one of operations that went wrong.

The first record of the set shows how far the second vocabulary goes. Its question is the shortest of the thirteen, and the compressed form turns it into a famous general form:

Here the compression does not remove a person, since there is none. It removes an object. The lost line says so in a sentence that would serve as a gloss on the move:

and the defect line names an operation that the compressed form performs on a single word:

Adolf Grünbaum, The Poverty of Theistic Cosmology (2004), asks whether the question of why anything contingent exists rather than nothing is a well-conceived question that calls for an answer at all, and sets out what it presupposes. The record does not go that far. It does not doubt that its compressed question is a question. But in this study's reading it does on a smaller scale something like what Grünbaum does: it treats the compressed question as something with presuppositions that can be named, and it names one, the running together of three senses of "reason". This study reads that defect line as the record set's most exact account of an idealization in Weisberg's sense: a distortion introduced on purpose, here into a word, and then written down.

The records whose questions are set in a concrete scene show the first vocabulary at its fullest. The hypothetical beginners' group under a cloudy sky asks the longest question of the thirteen, and its compressed form is among the most general:

"We" becomes "one"; "the stars" becomes "the object it hoped for"; "a good observation" becomes "something was gained". The compressed form is a question about inquiry as such, and it could be put about a failed experiment or an unanswered letter. What the record says it lost is the night itself:

The sentence names warmth and the length of a wait, and it does so as a loss the record admits rather than a detail it keeps. The defect line then warns in the opposite direction, against letting the concrete night stand in for a result:

Read together, the two lines of this record pull against each other, and the pull is the record's account of the compression. The general form lost the tea and the length of the night. But the record also says that the person's wish to call the night a good observation mixes two things that should stay apart. The record does not choose between them. It shows the compression as a trade in which the general question gains clarity about what counts as a result and loses everything about why the group went out.

The same trade appears where the setting is ordinary. The record on a planet where no life has been found loses, by its own account, the three things on which any decision about such a place would turn:

The last of these is worth a moment. Whether an act can be undone is a question the series puts to decision records, and no Cosmic Press record carries it as a field. Here it appears in the one place the form allows it: among the things the general question cannot hold. The record set knows, in this case at least, what a decision would need, and lists it as lost.

Rittel and Webber's distinction between wicked and tame problems is a lens here, in this study's reading, and a loose one: their paper is about problems of planning and social policy, which these questions are not. Read loosely, the compressed form is the tame version of a question, and the lost and defect lines are a list of what made it wicked. The record set's peculiarity is that it keeps both versions on the page.

1.3The top of the ladder

The six scale notes are the record set's second device. In the catalog's other topics they carry a personal question outward and then stop it. The sixth note of the catalog's first entry, an AI-generated question about rest and rent that this paper does not otherwise read, is typical in kind: it says that cosmic persistence alone cannot answer whether one may rest today. The top rung there works as a guard: it names the largest frame in order to say that the largest frame does not decide the small question. In the first 256 entries outside Cosmos, 153 of 243 sixth notes deny, limit or separate in this way.

In the Cosmos records, which are AI-generated and each set in a hypothetical scene, the question already stands at the top. A ladder whose last rung protects the person from the cosmos has nothing to protect when the question is about the cosmos. What the thirteen records do is turn the guard into a brake. Their sixth notes still deny, limit or separate (10 of 13 by the same test), but what they hold back is not the cosmos pressing on a person. It is the question pressing on what can be known. Their first notes almost never do so (1 of 13 by the same test). Instead, 8 of the 13 first notes speak of a feeling, a wish or an expectation, or of the evidence, fact or judgment it should be weighed against, and several of them keep the two apart. The ladder in these records is limited at both ends: at the bottom, in several records, feeling is kept from counting as evidence; at the top, the large question does not license a conclusion.

The record on loneliness shows the top-rung brake in its simplest form:

The sentence does not widen the question about loneliness to the cosmos. It narrows it: whatever the cosmos is like, the question cannot be answered from the absence of observation. The record on the Moon's resources ends the same way on a different subject:

and the record on the beautiful explanation ends with a sentence that could stand as a reading note on McAllister's subject:

James W. McAllister, Beauty and Revolution in Science (1996), studied how scientists' aesthetic preferences are influenced by the empirical track record of theories. In this study's reading, the record's question is the reverse case, an explanation with no track record because it concerns what cannot be observed, and the record's sixth note refuses to let beauty fill the gap. The record does not take McAllister's further step of explaining where aesthetic preferences come from. It does ask the question in its counterquestion, as Section 1.4 shows.

The first record of the set carries the brake at both ends, and its two outer notes are worth quoting together:

At the bottom rung, the person's feeling is kept from counting as evidence. At the top rung, two kinds of explanation are kept from standing in for each other. Between them the ladder runs through culture, traditions of knowledge, life and the Earth, and two of the middle notes also set things apart: the reason life arose from the reason for all existence, and the origin of the Earth from the ground of existence. The ladder here is not a zoom from the near to the far. It is a series of partitions.

One record shows what the whole ladder looks like when it is used this way from bottom to top. The hypothetical meeting with an intelligence that does not take questions as questions is the record most directly about the form of a question itself, and its six notes read as a single argument:

Each rung removes one assumption the asker brought. The first sets aside the expectation of an answer; the second the reading of silence as hostility; the third the idea that the asker's own customs are the only human ones; the fourth the equation of signal and question; the fifth the hope that shared ground guarantees shared expression; the sixth the premise that human questions are universal. By the top of the ladder the question of where conversation would begin has been stripped of every support it started with. The record's context says the scene is an artificial-intelligence thought experiment and not evidence of contact, and the ladder treats it accordingly: as an exercise in removing premises, not as a forecast.

Levin's argument that the observer imposes a perceptual bias, a filter through which a system is seen, fits this use of the ladder better than it fits the catalog's usual one. The usual ladder treats the six scales as levels of one world, each larger than the last. The Cosmos ladder, in this study's reading, treats them as six filters, each with its own blind spot, and the notes record what each filter cannot show. Marston's argument fits less well. The ladder is plainly constructed, as the studio's fixed form, but the thirteen records never ask who built it or whose activities it leaves out.

Guy Kahane, "Our Cosmic Insignificance" (2014), gives the sharpest contrast. It takes the view from the largest scale to be a question about significance: whether the vastness of the universe makes human lives small. The Cosmos records have some of the material for that question. This study reads the record on making what will vanish as the nearest to it, and its sixth note answers in the brake's manner:

The note neither affirms nor denies cosmic insignificance. It denies that the cosmic scale settles the matter automatically, which is a different and narrower claim. Kahane argues that worries about cosmic insignificance make sense and that the impression of insignificance is mistaken, but not for the usual reasons. The record does not take part in that argument. It stops a step earlier, at the claim that the top rung is not a standard of judgment. In this study's reading, the record set's ladder is built to keep the cosmic scale from deciding anything, and that is why its top rung, faced with a question about the cosmos, can only limit.

1.4Questions that answer questions

Every record, AI-generated and set in a hypothetical scene, carries a counterquestion. The form guarantees it, so its presence is not a finding. What the records put in it is. Of the 13 counterquestions, 8 address the asker or the premise directly; in the English, they speak of "we", "us", "my", "those who" or the thing "itself". All 13 end in a question mark, and none names a condition that could be observed and would reopen the question. In a decision record, the nearest field would say: if this happens, decide again. In the Cosmos records the counterquestion says instead: before deciding, ask what the question assumed.

The record on the family that must travel to see the stars has the counterquestion that moves furthest from its question:

The question was about whether the night sky is open to everyone. The counterquestion brings in people the question had not mentioned, whose interest runs against a darker sky. It does not reopen the question under a new condition. It shows that the question had a premise, that darkness is a good, which other people need not share.

The record on the unexplained signal turns the record's own caution against itself:

This last counterquestion is the closest the set comes to a decision rule, and it is still a question. The record's compressed form asked what verification the speed of making a discovery public should wait for. The counterquestion asks what the waiting does to who may judge. Haqq-Misra and others, The Benefits and Harm of Transmitting into Space (2013), hold that uncertainty about the outcome of contact makes it difficult to assess whether to engage in long-term, large-scale messaging to extraterrestrial intelligence; their subject is close to the record on the greeting, and the counterquestion of that record asks the opposite of caution:

Read against the work, the record holds the difficulty open from the other side: the work finds the decision hard because the outcome is uncertain, the record finds it hard because consent cannot be complete. Neither the record nor this study says what should be done.

The record on the beautiful explanation sends its counterquestion to the aesthetic judgments scientists pass on their theories, the subject of McAllister's book (this study does not claim the book answers it):

and the first record of the set puts to its compressed form a question about the form's own presupposition, which this study reads as the kind of question Grünbaum's paper asks of the Leibnizian question (the work's note does not say what it concludes about this one):

These two show the limits of the device. A counterquestion can point at a premise; it cannot test one, and the records give no way to tell whether a premise has been found wanting.

1.5What this chapter adds

This chapter has argued three things about thirteen AI-generated records. The compression removes the asker, and the record set says so in two accounts, one of what was lost and one of what the general form does wrong. The ladder of scales, which in the catalog's first part stops the cosmos from deciding a personal question, becomes in the Cosmos records a brake on the cosmic question itself, limiting what can be concluded at the top and what can count as evidence at the bottom. The counterquestion does not reopen a question under a condition; it exposes a premise. To the debate on formulation the chapter adds a record form that formulates twice and writes down the difference; to the debate on idealization, an idealization that names its distortions but does not correct them; to the debate on scale, a ladder whose top rung, when the question is already at the top, can only limit.