Open Problems in the Jurisprudence of Generativity
Abstract
Companion work develops a framework in which value is generated by a
relational configuration rather than by any party to it, and is therefore not
decomposable in advance into shares belonging to parties who existed before the
configuration did. Applying that framework to questions of ownership,
transfer, and remedy produces a determinate negative result and then stops.
This paper states where it stops. Twenty-one problems are set out, each with
the argument that produces it, the downstream questions it renders
unanswerable, and what would defeat it. The problems are not independent. One
of them asks whether a generative capacity and the conditions of its exercise
can be separately identified at all, and if that question has no affirmative
answer then six of the others are not problems in their own right but
appearances of it. Two further problems are independent of that question and
survive any answer to it: that a capacity transferred without diminution defeats
the exclusion and depletion premises of the property model, and that comparing
what two holders would generate requires a measure across kinds that may not
exist. Three entries are marked, and the marks matter. One records an error in
the author’s own reasoning rather than a difficulty in the subject, one records
a repair that was proposed and then defeated, and one records the single
question the framework has settled, included so that the list cannot be read as
uniformly negative. The paper claims that stating this structure is a result.
A framework that returned confident answers here would be one that had smuggled
back in the decomposition it set out to deny, and the strongest reason for
setting the problems out in order is to make that smuggling visible when it
occurs.
Keywords: generativity; property; inalienability; individuation;
Discussion Paper Note
This paper is a preliminary discussion paper intended to share an evolving idea
and invite further dialogue, criticism, revision, and independent development.
Its problems, arguments, and structural claims remain provisional. Circulation
across scholarly and practical communities is part of the purpose of releasing
the manuscript at this stage.
The author treats the viewpoints, concepts, and lines of reasoning presented
here as contributions to a shared field of inquiry. Similar or related ideas
may have appeared in other intellectual, cultural, and disciplinary traditions,
and Section 14 states those the author is aware of. The
manuscript separates the researcher-origin proposal from later formal
reconstruction and leaves historical priority open pending a systematic
originality review.
Readers are encouraged to question, test, revise, extend, reinterpret, or
independently develop the material presented here. A problem list invites
solution more directly than an argument does, and the author would regard any
entry below being closed as the best possible use of the document. Where
appropriate, acknowledgment of this paper as one point of encounter in the
development of a related idea is appreciated. Such acknowledgment records an
intellectual route; the ideas themselves remain available for criticism,
revision, and independent development.
Responsible Use and Rights Reservation
This section separates requested scholarly conduct from the legal permissions
stated on the following page.
One caution is specific to this manuscript. A list of unsolved problems can be
used to argue that nothing can be decided, and a party that benefits from a
present arrangement may find that conclusion convenient.
Section 13 therefore states what the framework does
settle, and Problem ? records an occasion on which the
author himself drew an unwarranted conclusion in favour of an existing holder.
The author asks that the document be read with that entry in view, and that no
entry below be cited as showing that a question has no answer when what it
shows is that the author has not found one.
The author encourages good-faith discussion, criticism, independent inquiry,
and responsible use of the material in this work. Separately from the licence’s
terms, the author asks users to consider foreseeable harms when adapting or
applying the material. This ethical request leaves the licence’s permissions
and legally authorized uses unchanged.
The author retains the rights preserved under CC BY-NC 4.0 and may pursue
remedies to which the author is legally entitled for breach of the licence or
violation of the author’s independently applicable rights. Reuse remains
independent from authorial endorsement. Third-party rights require authorization
from their respective holders where applicable. Copyright exceptions and
limitations, including applicable forms of fair use or fair dealing, remain
fully available.
Notices
Status.
This working draft records an evolving stage of the author’s position and is
circulated for discussion. The problems, their numbering, and their dependency
structure remain subject to revision. Formal treatment of any individual
problem, empirical work, and an originality audit remain future research
stages.
Licence.
Except where otherwise indicated, copyright 2026 Wanhong Huang. This work is
made available under the Creative Commons Attribution-NonCommercial 4.0
International License (CC BY-NC 4.0). Subject to its terms, the licence permits
sharing and adaptation for noncommercial purposes with appropriate attribution,
a link to the licence, an indication of changes, and attribution that preserves
the licensor’s independence from the reuse. The licence deed and legal-code
link are available at
https://creativecommons.org/licenses/by-nc/4.0/. The licence governs in
case of conflict with this summary. Third-party material remains subject to the
rights held by its respective rights holders.
Statement on the use of language models.
The exploratory discussions and preparation of this paper involved Anthropic’s
Claude and OpenAI’s ChatGPT. These systems supported exploratory dialogue,
literature and legal survey with subsequent verification against publisher,
journal, court, and institutional pages, argumentative criticism, and drafting
in . Several of the problems below were arrived at by a process in
which a proposed resolution was advanced and then defeated, and the record of
those defeats is retained in the text where it is informative. The author
selected the research questions, directed and approved the theoretical
commitments and epistemic status of the claims, and bears sole responsibility
for the manuscript. Every source cited here has been verified against a
publisher, journal, court, or institutional page; sources encountered in survey
and not so verified are excluded from the manuscript and recorded separately in
the accompanying literature audit.
Companion papers.
This paper states the unsolved remainder of a programme whose settled portions
appear elsewhere: Relational Value Exploitation: Taxonomy and Mechanisms,
which supplies the lifecycle, the aspect grammar, and the corpus of sites
referred to below; Relational Colonization: Power, Grammar, and the
Reconfiguration of Relational Spacetime, which supplies the grammar and
closure apparatus; and two papers developing a gauge-theoretic account of value
circulation, The Geometry of Value and The Structure of
Exploitation on Heterogeneous-Value Loops.
Suggested citation.
Huang, Wanhong. “Open Problems in the Jurisprudence of Generativity.”
Working discussion paper, 2026.
1. Introduction
This section states the object of the paper, the reasons for presenting a
problem list as a result, the admission standard each entry meets, the meaning
of the three marked entries, and the scope of what follows.
Companion work develops a framework in which value is generated by a relational
configuration rather than by any party to it. Its founding commitment holds
that such value has no decomposition, in advance, into shares belonging to
parties who existed before the configuration did. Applied to questions of
ownership, transfer, and remedy, that commitment yields determinate negative
results and then reaches its limit. It establishes why certain allocations are
unavailable. It leaves open what should be done in their place, and on several
questions it leaves open whether the question is well formed. The present paper
states where the framework reaches its limit, and states it as an ordered
structure of twenty-one problems, each carrying the argument that produces it,
the downstream questions it renders unanswerable, and the finding that would
close it. The entries are ordered by dependency.
Three considerations support presenting such a list as a result, of which the
third is operative. The first is that the entries are derived. Each is
accompanied by an argument, and each stands or falls with that argument.
Problem ? argues that the portion of a generative capacity
which makes it generative rather than productive lies outside the evidence by
construction. Problem ? argues that attribution requires a
separation between a capacity and the conditions of its exercise which
co-evolution withholds. A reader who rejects an entry rejects an argument.
The second consideration is that the dependency structure carries information.
Twenty-one unrelated difficulties would constitute a scattered field.
Section 6 argues that these compose a single trunk,
descending from one question about the existence of the object, together with
branches that survive any answer to that question. The structure indicates
where work is productive and where it is redundant.
The third consideration is the operative one. The framework’s central
commitment is a denial of decomposability, so any proposed solution to the
allocation problems below requires examination for whether it has reintroduced
a decomposition under another name. A confident answer in this region
constitutes prima facie evidence of such a reintroduction, and a problem list
stated in dependency order is the instrument by which the reintroduction
becomes visible. The list serves as a record of what remains unknown and
equally as a test that proposed answers have to pass.
Four conditions govern admission to the list. An entry is derived, in the sense
just given. An entry states something shown to be unavailable, or a question
shown to be ill posed, or a proposition whose truth value remains undetermined;
a call for a theory of some subject falls outside this standard. An entry names
what fails downstream if it stands, which is the feature that makes the list
usable. And an entry states what would close it, so that an entry for which the
author can imagine no settling finding counts as a posture and is excluded.
Three entries carry marks, and the distinctions among the marks are
load-bearing. Problem ? records an error in the author’s own
reasoning. Having argued that the unpredictability of future generation defeats
a proposal to reallocate a capacity, the author took the same unpredictability
to support leaving the capacity with its present holder. That inference fails,
and it contradicts a position the framework holds elsewhere; the entry retains
the error in its statement. Problem ? records a repair
proposed and then defeated. Confronted with the unavailability of
counterfactual generation, the author proposed that legal certainty
requirements measure at the wrong level, and that claims about the persistence
of a grammar remain available where claims about its products do not. That
repair fails for the reason given in the entry, and the entry retains it
because the manner of the failure locates the difficulty more precisely than
the original statement does. Problem ? states a settled result
and appears in the numbered sequence so that the list resists reading as
uniformly negative. A document recording only the difficulties of a subject,
and neither its author’s errors nor his settled conclusions, offers a reader no
means of calibrating the remainder.
Three exclusions bound the paper. It solves no entry, and where the author
holds an inclination the entry marks it as an inclination. It supplies no
normative theory, and the normative entries in Section 12 appear
as open. It takes no position on any contested question of politics,
sovereignty, or comparative ethics, and contains no example drawn from such a
question. The generative artificial intelligence material used at
Problem ? and elsewhere appears because that setting renders
an abstract structure enumerable, and carries no view about the industry.
Section 5 fixes the objects the problems range over.
Section 6 states the dependency structure and the format of
an entry. Sections 7 to 12 give the entries in
five layers. Section 13 states the results the framework
settles, Section 14 the traditions in which several entries
have prior statements, and Section 15 the limits of the document,
its intended uses, a summary table, and the conclusion.
2. Preliminaries
This section fixes the commitment on which every entry rests, the four objects
the entries range over, an illustration in which those objects come apart
cleanly, and one result carried forward from companion work as a premise.
Nothing in this section is argued; the companion papers supply the arguments,
and the statements here are given in the form the later entries use.
2.1 Commitment on the composition of relational value
Relational value is value generated by a relational configuration rather than
by any party to it, and is accordingly not decomposable in advance into shares
belonging to parties who existed before the configuration did.
Trust, shared language, coordination routines, tacit knowledge, standing, and
the capacity to generate again are all of this kind. Definition ?
is the source of every negative result in this paper. Where an allocation
requires a decomposition, the commitment denies that the decomposition exists,
and the allocation fails.
2.2 Objects of the analysis
The problems below turn on distinctions among four things that are routinely
conflated.
A product is something a configuration has generated. Products are
individuable, datable, and in the ordinary case ownable.
The generative capacity $\gen$ of a configuration is its capacity to
generate value that has not yet been generated.
The grammar $\gram$ of a configuration is the structure that fixes which
interactions are available within it. Interactions realized under a grammar are
distinct from the grammar itself, so a position may operate on $\gram$ while
performing no interaction.
The conditions $\cond$ of a configuration are what has to obtain for its
grammar to be exercised: resources, infrastructure, standing, time, and the
continued availability of the parties.
Two further notations recur. $\opt_t$ is the space of relations available to a
party at time $t$; a contraction of $\opt$ is the formal shape of a foreclosure.
$O_t$ is a product generated at $t$.
The relation ordinarily assumed between these objects is written at
Equation (1), in which a capacity is exercised under conditions to
yield a product.
$$O_t = \gen\bigl(\cond_t\bigr)$$
Problem ? asks whether Equation (1) is well
formed, and every problem in Sections 8 and 9
presupposes an affirmative answer to that question.
2.3 Separation of the objects in generative artificial intelligence
Generative artificial intelligence separates the four objects with unusual
clarity, and the separation is worth stating once here because it recurs.
A provider holds a model. A user contributes an interaction history. Products
are generated. Under the terms of major providers, whatever rights the provider
may hold in the outputs are assigned to the user, and the assignment is
qualified: the operative language in the terms of several providers assigns
such rights “if any”
(Henderson, 2025). That qualification is not
drafting caution. In the United States, the human authorship requirement means
that material generated without sufficient human contribution is not
copyrightable at all
(Circuit, 2025; Office, 2025), so a contractual assignment may
operate over subject matter in which no property right subsists.
The user therefore holds products and holds nothing else. The model, the
infrastructure, the memory mechanism, and the terms on which future generation
may occur are all held by the provider. What the user and the model jointly
constitute over a long history is a fifth thing, and that fifth thing is the
object of this paper.
The example is used throughout because it makes an abstract structure legible
and because its components can be enumerated, not because the paper takes a
position on the industry. Problem ? states what is peculiar
to it. Everything else is general.
2.4 Result carried forward from companion work
One conclusion from the companion work is used as a premise below rather than
re-argued, and appears here because several problems depend on it.
That a party brought a configuration into being, or contributed to the
constitution of a party who later enters other configurations, grounds no claim
over what those configurations subsequently generate. The argument is that any
contrary principle fails to terminate, since every configuration has
constitutive ancestry without limit, and that the resulting claims would be
collected by whoever is positioned to press them rather than by whoever is
entitled.
Settled Result ? is applied in both directions throughout, and
Problem ? records why its symmetry has to be defended against
erosion.
3. Architecture of the Problem List
This section states the dependency structure among the twenty-one entries, the
arrangement of those entries into five layers, the format in which each entry
appears, and a caution governing the use of the list. The dependency structure
constitutes the paper’s principal structural claim, and it holds that the
entries compose a connected structure rather than a set of independent
difficulties.
3.1 Dependency structure
Problem ? asks whether a generative capacity and the
conditions of its exercise can be separately identified. If that question has
no affirmative answer, then Problems ? through
? are not problems in their own right; they are
appearances of it under different descriptions. Two further problems are
independent of it and survive any answer: Problem ?,
concerning transfer without diminution, and Problem ?,
concerning value across kinds. The normative problems of
Section 12 are likewise independent.
The practical import of Claim ? is a research instruction.
Work on identification, on boundaries, on the identity conditions of a grammar,
and on partial transferability may all be work on a single question that has
not been posed. A solver’s time is better spent on
Problem ? than on any of its descendants.
The two independent branches are worth naming for the opposite reason. They
cannot be dissolved by any progress on the trunk, and they are the entries most
likely to be closed by ordinary means, since neither requires an answer to the
ontological question.
3.2 Arrangement by layer
Table 1 gives the arrangement. The layers are ordered by what
each presupposes: a layer cannot be worked on coherently unless the layers
above it are either answered or explicitly bracketed.
| @P0.09P0.21P0.15Y@
| Layer | Question | Problems | Presupposes |
|---|---|---|---|
| 0 | Does the object exist as a persisting thing? | ?, ? | Nothing. |
| 1 | Can it be known? | ?–? | That there is something to know, hence Layer 0. |
| 2 | What are its boundaries? | ?–? | That it can be identified, hence Layer 1. |
| 3 | Can it be allocated? | ?–? | That it has boundaries, except for the two independent entries. |
| 4 | Can loss be attributed? | ?–? | That allocation questions have determinate objects. |
| 5 | What is owed? | ?–? | Nothing above; independent. |
Table. Layers of the list and their presuppositions
3.3 Format of an entry
Each entry has four parts.
The statement gives the problem in a form that could in principle be
answered.
The derivation gives the argument that produces it. This is the part
that carries the weight, and a reader who wishes to reject an entry should
engage here.
The consequences name what fails downstream if the entry stands.
The defeater states what would close it. Where the author regards a
defeater as likely, the entry says so.
3.4 Caution on the use of the list
A list of open problems can be deployed to argue that nothing may be decided,
and the party favoured by an existing arrangement will find that argument
convenient. Two features of the document are intended to prevent it.
Section 13 states what the framework does settle, and one
settled result appears within the numbered sequence.
Problem ? records an occasion on which the author drew
precisely the convenient conclusion and was wrong to do so. The inference from
“neither party can prove its case” to “the present holder retains” requires
a premise the framework denies elsewhere, and the entry preserves the error
rather than the correction alone, because the error is the more instructive
object.
4. Layer 0: Existence of the Object
This section states the two entries concerning the existence of a persisting
object. Its role in the paper is foundational: an unfavourable answer to either
entry alters the status of most later entries, per Claim ?.
The method is analysis of the standard functional form in which a capacity is
related to its conditions, followed by examination of what survives if that
form fails. Problem ? concerns the separate identification
of a capacity and the conditions of its exercise;
Problem ? concerns what remains available if such
identification fails.
Can a generative capacity $\gen$ and the conditions $\cond$ of its exercise be
separately identified?
Derivation.
Equation (1) writes a product as a capacity exercised under
conditions, and in doing so treats $\cond$ as exogenous. That treatment is what
licenses the counterfactual on which almost every legal and normative question
in this area depends: what this capacity would yield under other conditions.
The treatment is not obviously correct. A capacity is exercised, and the
exercise alters the conditions under which the next exercise occurs. A
laboratory that publishes acquires standing, which alters its access to
resources, which alters what it can attempt next. A user who converses with a
model accumulates a history, which alters what the next exchange can be built
on. The relation is at least as well described by
Equation (2) as by Equation (1).
$$\bigl(\gen_t, \cond_t\bigr) \longmapsto \bigl(\gen_{t+1}, \cond_{t+1}\bigr)$$
Under Equation (2) the question “what would $\gen$ yield under
$\cond’$” has no determinate referent. The history that would have produced
$\cond’$ rather than $\cond$ is the same history that produced $\gen$; a world
in which the conditions differ is not a world in which the same capacity meets
different conditions, because the capacity was formed by the conditions it met.
The counterfactual asks us to hold fixed a term absent from the alternative
history.
Consequences.
By Claim ?, Problems ?
through ? presuppose an affirmative answer here. If
there is no answer, those six are appearances of this one rather than problems
in their own right, and work on them separately is misdirected.
The consequences reach further than the list. Attribution of loss requires the
separation, as Problem ? sets out. Any statement of the
form “the capacity survived but the conditions changed” requires it. So does
its negation. The three descriptions that Problem ? shows to
be observationally indistinguishable are not merely hard to tell apart; if this
problem has no affirmative answer, the distinction among them is not
well formed.
Defeater.
An argument that the coupling between $\gen$ and $\cond$ is weak enough over
relevant intervals that the exogenous treatment is a good approximation would
close this, and would close it for practical purposes even if strictly false.
That is the author’s expectation of how it will be closed, if it is. The
argument would need to specify the interval and the tolerance, and to say what
happens at the boundary, since an approximation that fails exactly where
intervention occurs is not useful.
A second and less likely defeater would be a demonstration that some component
of $\gen$ is genuinely condition-independent, so that the object splits into a
portable part and a coupled part. The author regards this as unlikely for the
same reason that Problem ? is hard: partitions of this kind
tend to be stipulated rather than found.
If $\gen$ and $\cond$ are not separately identifiable, is there some persistence
weaker than identity that nonetheless makes statements about a configuration’s
continued generating true or false?
Derivation.
Suppose Problem ? has no affirmative answer. Then there is
no object that persists across time in the way ownership requires, and the
failure of the property model is immediate. But the failure does not stop
there. “This laboratory has been producing good work for a decade” would also
fail, and so would “this relation is no longer generating what it did.” These
statements plainly have content. People assert them, act on them, and are
sometimes shown to be wrong about them, which is the ordinary mark of a claim
with a truth value.
Two positions are available and the author cannot choose between them.
On the first, persistence obtains in a form weaker than identity. Something
survives that is not a self-identical object: a continuity, a lineage, a
characteristic manner. Statements about continued generating are true in virtue
of it. The task is then to say what it is, and to say why it supports the
statements it does without supporting ownership.
On the second, no such thing survives, and our language systematically posits an
object here that is not there. The positing is useful, which is why it is
stable, but claims resting on it inherit its status. The task is then to explain
the utility of a fiction and to say which practices may safely rest on it.
Consequences.
The two positions lead to different inquiries and different papers. On the
first, Problem ? is the central question and there is an
answer to be found. On the second, Problem ? is malformed and
the correct response is to reconstruct the practices that presuppose it.
Defeater.
A criterion of continuity for a generating configuration, weaker than identity,
which supports the ordinary assertions and does not support ownership. The
author does not have one. He notes that the requirement is demanding in a
specific way, being strong enough to make “still generating” true and weak
enough that “therefore ownable” does not follow, and criteria proposed
for other purposes tend to fail one or the other.
5. Layer 1: Epistemic Access to the Object
This section states four entries concerning epistemic access to a generative
capacity. Each presupposes an affirmative answer to
Problem ?, so the section proceeds on the assumption that
a capacity exists and admits description apart from its conditions. The method
is to ask, for each of four objects of knowledge, what a record of past
generation establishes about it: the grammar of a configuration at
Problem ?, the unexercised range at
Problem ?, generation under an alternative grammar at
Problem ?, and the trajectory of a capacity at
Problem ?. The entries are ordered so that the third follows
from the second.
Can the grammar of a configuration be identified from the products it has
generated?
Derivation.
A grammar is a structure that produces an open set of admissible instances. Any
finite set of instances is therefore compatible with many grammars, and the
compatible grammars agree on the observed instances by construction.
The difficulty is not that the compatible grammars are numerous. It is where
they differ. They agree on what has been produced and diverge on what has not,
which is to say they diverge on everything a claim about generativity concerns.
Observing that a configuration has produced $a, b, c, d, e$ under conditions
$\cond$ narrows the field of candidate grammars but leaves them free wherever
$\cond$ has not obtained.
The resulting knowledge has a definite shape. It supports extrapolation along directions already
realized, and lapses laterally. That a group has
worked productively in one area under one arrangement gives some ground for
expecting further work in that area under that arrangement. It gives
essentially no ground for any expectation about a different area under a
different arrangement, because the candidate grammars consistent with the
record differ precisely there.
Consequences.
Every downstream question that requires knowing what a configuration’s grammar
permits, rather than what it has done, inherits this. Two are worth naming.
Problem ? asks when a grammar has changed. That question
presupposes that the grammar can be identified at two times and compared. If
identification is local in the sense above, the comparison is available only
within the region where conditions have held constant, which is the region
where nothing interesting has happened.
Problem ? records a proposal that claims about a grammar’s
persistence are available where claims about its products are not. This problem
defeats that proposal. Identifying a grammar is harder than predicting a
product, not easier, because a product is a single thing and a grammar is a
class of them.
Defeater.
An identification result: conditions under which a finite product set,
perhaps with structure or with active probing, determines a grammar up to some
equivalence relevant to the questions asked here. The author notes that results
of this general kind exist in the study of formal learning, that they typically
require either negative examples, or a restriction of the hypothesis class, or
a query mechanism, and that the availability of any of these for a social
configuration is exactly what would need to be shown.
Is the part of a generative capacity that makes it generative, rather than
productive, available to observation in principle?
Derivation.
A capacity has an exercised portion and an unexercised range. The exercised
portion is a record of production: what was made, when, by whom, under what
conditions. It is observable, and in a well-documented case it is observable in
detail.
The unexercised range is what distinguishes a generative capacity from a
productive one. A capacity whose outputs can be specified in advance is a
production function; what makes a capacity generative is that it yields what
has not been specified. That range has never entered the record, because
entering the record is what exercise consists in.
A related asymmetry is established for knowledge: a knower may know more than
the knower can tell, so that what is possessed exceeds what can be articulated
(Polanyi, 1966). The present claim is adjacent and distinct. There the
gap is between possession and articulation, and the unarticulated portion is
possessed. Here the gap is between exercise and range, and the unexercised
portion has not been possessed in any sense that could be reported, since
possession of a generative capacity consists in what it would do rather than in
what its holder could say.
Hence the statement of the problem. The portion of a capacity that makes it
generative is, by construction, outside the evidence. This is not a limitation
of instruments and would not be relieved by better observation, because the
object is constituted by what has not been observed. Observation converts
generativity into production, one instance at a time, and what remains
generative is what has not been so converted.
Consequences.
Every legal or normative claim about generativity is a claim about the
unobserved range. That a configuration produced well is a claim about the
record. That it will, or would, or could produce well is a claim about what has
never been seen. The recurring finding that such claims cannot be adjudicated
is explained by this, and explained more precisely than by an appeal to
uncertainty about the future: the difficulty is not that the future is
uncertain but that the object of the claim is constituted by its
unmanifested range.
Problem ? is a corollary rather than an independent
entry. Problems ? and ? inherit the
difficulty in their respective forms.
Defeater.
An account of a property of a capacity which is observable in the exercised
portion and which entails something about the unexercised range. Dispositional
properties in other domains have this character: a solubility is observed in
particular dissolvings and is taken to license claims about occasions that
never arise, because a theory connects the two. The defeater here would be the
corresponding theory for generative configurations. Its absence is not obvious
as a matter of principle, which is the reason this entry is a problem rather
than a result.
Can it be established what a configuration would have generated under a grammar
other than the one it operated under?
Derivation.
This follows from Problem ? together with
Problem ?. The counterfactual concerns the unexercised range,
which is unobserved by construction; and it concerns a grammar other than the
one instantiated, about which the record is silent by the lateral argument
above. The entry is listed separately because it is the form in which the
difficulty reaches legal doctrine, where a claim that a party would have
generated some determinate thing is the ordinary vehicle for a claim of loss.
Consequences.
The companion work states that a configuration under an imposed grammar
generates a composition different from what it would otherwise have generated,
and that no party can demonstrate the difference. That proposition is an
implication of the framework and is not testable in the form given. The present
entry records why, and locates the reason one level deeper than the companion
paper does.
Defeater.
Whatever defeats Problem ? defeats this. Separately, a
domain in which comparable configurations under differing grammars are
available in sufficient number might support population-level counterfactuals
even where individual ones remain unavailable. That would relieve the
difficulty at the level of rules while leaving it untouched at the level of
cases, which is precisely the division that Problem ?
discusses.
Can the trajectory of a generative capacity be predicted, as distinct from the
products it will yield?
Derivation.
The two are different questions and the second does not reduce to the first.
Even granting a fully identified grammar at $t$, its successor at $t+1$ remains
open. Which generative capacities are stable across time, which depend on a
particular history and lapse when the history is interrupted, which transfer
across domains and which do not, are questions about the map in
Equation (2) rather than about any product.
Problem ? stands independent of Problem ?.
Identification concerns whether the grammar at a time can be recovered from the
record.
Evolution concerns whether, given the grammar at a time, its successor can be
derived. A solution to either leaves the other open.
The independence has a practical form. A long record of productive work in an
area under an arrangement licenses some expectation about continuation. It
licenses much less about the same configuration in a different area under a
different arrangement, and the reason is not only lateral ignorance of the
present grammar but ignorance of how grammars change when conditions change.
The second ignorance would remain even if the first were relieved.
Consequences.
Any regime of remedy directed at the preservation of a capacity requires a
judgement about whether a given intervention changed the trajectory. That
judgement requires knowing what the trajectory would have been.
Problem ? is the acute form: without an account of how
grammars evolve, there is no way to say whether an observed change is the
grammar developing or the grammar ending.
Defeater.
Empirical regularities in the persistence of generative capacity, of a
kind that would support statements about the class even without supporting
statements about any member. Such regularities may exist and would be
discoverable by ordinary means. This is among the more tractable entries in the
list, and the author expects it to be closed at population level well before
the entries above it are closed at all.
6. Layer 2: Boundaries of the Object
This section states four entries concerning the boundaries of a generative
capacity, presupposing the results of Sections 7
and 8. Three boundaries are treated in turn: the boundary among
parties at Problem ?, the boundary across time at
Problem ?, and the boundary between what moves and what
remains under a transfer at Problem ?.
Problem ? closes the section by separating five
operations that the word transfer conflates. The method throughout is to
identify the determination each downstream claim presupposes and to ask what
supplies it.
Which parties belong to a generating configuration?
Derivation.
Relational value is generated by a configuration, so every claim about it
presupposes an answer to which parties compose the configuration. The
presupposition is rarely examined, and frequently constitutes the whole of what
is in dispute.
Two ordinary cases show the shape. A family may or may not be a member of the
household its adult child forms; whether it is settles whether subsequent
arrangements are internal governance or external interference, and the parties
disagreeing about the arrangement are ordinarily disagreeing about the
membership question without saying so. In generative artificial intelligence,
whether the provider is a party to the configuration a long-standing user has
built, or an external supplier of infrastructure to it, settles what obligations
the provider bears, and no source of an answer is available.
Membership stands as a determination made rather than a background fact
awaiting discovery. Deciding it is an
exercise of the power the companion work names as schema power, since a
position able to determine who counts as a party thereby determines against
whom claims may be made and by whom. That makes the membership question the
most upstream position in the apparatus and simultaneously the least settled.
Consequences.
Every allocation problem in Section 10 presupposes a settled
membership. So does every attribution problem in Section 11, since
a loss is borne by a party and the party has to be inside. So does the
application of the companion work’s four admissibility conditions, each of
which is applied to a named position within a configuration whose extent is
assumed.
Defeater.
A criterion of membership that is not itself an exercise of the power it would
regulate. The obvious candidates fail in an instructive way. Contribution as a
criterion collapses into the origin question closed at Settled
Result ?. Consent presupposes an already-constituted party
competent to give it. Dependence would make every supplier a member. The
difficulty is that membership criteria are ordinarily supplied by one of the
parties, and this problem is the general statement of why that is a defect
rather than an administrative convenience.
Under what conditions is a grammar at $t+1$ the same grammar as the grammar at
$t$?
Derivation.
Any obligation not to destroy a generative capacity requires a distinction
between destruction and development. Without it, a party that has altered a
configuration can always characterize the alteration as evolution, and a party
complaining of loss can always characterize development as destruction. Neither
characterization is answerable.
The framework as it stands offers scalar quantities. It has a direction of
movement for $\opt$, the space of relations available to a party. It has an
increase or decrease in $\gen$ across a cycle. Neither is adequate here,
because a grammar is not a scalar. Two grammars may license the same number of
interactions and license entirely different ones, and the question of identity
concerns which paths remain available rather than how many.
What is needed is structural: an account of which generative paths persist,
which lapse, and which appear, together with a criterion for how much of that
structure may change while the grammar remains the same one. The framework has
no object of this type.
Consequences.
Problem ? is a special case: to say that a transfer preserved
some fraction of a capacity presupposes that the transferred thing can be
compared with the original, hence that identity conditions exist.
The consequence for remedy is severe. A continuity obligation, if one existed,
would be an obligation not to destroy. Its content depends entirely on this
problem, and in its absence such an obligation has no determinate content even
where all parties agree that something has been lost.
Defeater.
Identity conditions for structures of this kind, presumably weaker than
isomorphism and stronger than mere continuity of the parties. Related notions
exist in the study of dynamical systems, where an account of when a system has
undergone a qualitative change rather than a quantitative one is available.
Whether that account transfers is the question, and the author regards this
entry as the most likely of the structural problems to be closed by importing
apparatus from elsewhere.
What is the status of a capacity that is transferable in part?
Derivation.
The framework treats separability as binary. A capacity is either severable
from its configuration or not, and the companion argument proceeds by showing
that transfer destroys what it purports to move.
Actual cases are graded. Documented methods, institutional forms, corpora,
tooling, and codified procedure move; standing, coordination, and the tacit
component do not. What arrives at the destination is neither the whole nor
nothing. The literature on the transfer of practice within organizations
records exactly this: transfer is impeded rather than impossible, and the
impediments admit degrees
(Szulanski, 1996). The theory of routines
as organizational memory holds that routines are constitutively organizational,
which is a claim about the residue rather than about the whole
(Nelson, 1982), and rests in turn on the observation that skill
exceeds what its holder can state (Polanyi, 1966).
A graded phenomenon requires a graded treatment, and the framework supplies
only the binary one.
Worse, the natural way to state a grade presupposes Problem ?.
To say that forty per cent arrived is to compare what arrived with what was
there, which requires identity conditions for the compared objects. Absent
those, the fraction has no denominator.
Consequences.
This is a defect in the framework rather than a difficulty in the world, and it
is recorded as such. The companion argument that severance destroys the object
holds for the residue and overstates for the transferred part. A legal regime
built on the binary claim would misdescribe the ordinary case.
Defeater.
Closing Problem ? would supply the denominator and reduce this
to a measurement question. Alternatively, a demonstration that the transferable
component is never the generative component, so that what moves is always
product and never capacity, would restore the binary treatment. The author
regards the second as unlikely: a codified method is not merely a product, and
it does bear on subsequent generation.
Which of several distinct operations does “transferring a generative capacity”
denote?
Derivation.
At least five operations are run together by the phrase.
Capacity transfer: the receiving party comes to be able to generate what
the originating party could generate. Control transfer: authority over
the conditions of exercise moves, while the capacity stays where it was.
Claim transfer: a right or interest moves, with no change to any capacity
or condition. Access transfer: the receiving party gains the ability to
invoke a capacity that remains constituted elsewhere. Relational
transplantation: the configuration itself is relocated, with its parties, into
a different setting.
The five have different truth conditions and different consequences. Access
transfer is the ordinary case in service arrangements and involves no movement
of anything. Control transfer is what occurs when a party acquires the
infrastructure on which a configuration depends. Relational transplantation is
what an acquisition of a team attempts.
The equivocation is not merely terminological, because a familiar dispute
depends on it. Where one party learns another’s method, whether this is
transfer or replication has no answer until the sense is fixed: it is
replication under capacity transfer, since the originating party retains what
it had; it may be nothing at all under claim transfer; it is not
transplantation, since no configuration moved.
Consequences.
Arguments in this area proceed by equivocation more often than by
disagreement. A claim that generativity cannot be transferred, if it is a claim
about relational transplantation, is compatible with the routine availability of
access and claim transfer; and a claim that it obviously can be transferred is
frequently a claim about access.
Defeater.
This entry is closed by stipulation rather than discovery, and the author
regards it as the most easily closed on the list. It is included because the
stipulation has not been made, and because several disputes dissolve once it is.
7. Layer 3: Allocation of the Object
This section states five entries concerning the allocation of a generative
capacity among parties. Two of the five, Problems ?
and ?, stand independent of every preceding entry and
survive any answer to Problem ?; the section marks them as
independent where they appear. The method is to take the standard analytic
apparatus of ownership, apply it to a generative capacity, and identify the
components that fail. Problem ? examines the exclusion and
depletion premises; Problem ? examines the four available
allocation rules; Problem ? examines the strongest objection to
the preceding entry; Problem ? examines the measure a
comparative argument requires; and Problem ? examines what
follows where neither side can carry its case, and records an error the author
made at that point.
If a capacity can be conveyed to a second party without diminishing the first
party’s capacity, does the property model apply to it at all?
Derivation.
The apparatus of ownership presupposes two features of its object that need not
hold here.
The first is exclusion. The analysis of ownership into incidents begins with a
right to possess and proceeds through rights to use, to manage, and to the
income and capital of a thing (‘e, 1961); each is stated
against a background in which one party’s holding is another’s exclusion.
Property rights are additionally in rem, imposing an obligation on an
indefinite class of strangers, and the standardization of property forms is
explained by the information costs that obligation imposes
(Merrill, 2000). Both features assume that there is something a
stranger can be excluded from.
The second is depletion. Where a thing passes from one holder to another, the
first ceases to have it. The remedial apparatus follows: conversion, restitution,
and damages all measure what the transferor lost by the transferee’s gain.
A capacity conveyed by demonstration or instruction has neither feature.
The originating party retains what it had. Nothing was subtracted, and no
stranger was excluded from anything. The recipient now generates, and the
originator generates no less.
The observation is not new for information goods generally, and the treatment
of intangibles has long accommodated non-rivalry. What is not settled is
whether a capacity behaves like an information good in the relevant respect. A
codified method may be copied and both parties hold it. A capacity that
consists in the coordination of particular parties may not be copyable at all,
as Problem ? records. So the entry is not that generativity is
non-rival, but that its rivalry is undetermined and that the property model
presupposes an answer.
Consequences.
This entry is independent of the whole trunk. Even granting that a capacity is
separable from its conditions, identifiable, and bounded, the property model may
fail to apply for a reason that has nothing to do with any of that. It is
therefore among the strongest available grounds for the framework’s negative
conclusions, and the author notes that he has relied on weaker ones.
Defeater.
A demonstration that the generative component is rivalrous even where the
codified component is not, so that what genuinely passes is always the
non-generative part. This would restore exclusion and depletion to the object
that matters and would also close Problem ? in the same stroke.
The author regards it as the most consequential single result that could be
obtained anywhere in this list.
Do the available allocation rules fail independently, or do they fail for a
single reason?
Derivation.
Four rules are available for allocating an interest in a jointly generated
capacity, and each is known to fail.
No ownership. The capacity is left unallocated. This is coherent, and it
supplies no protection to any contributor and no ground for any complaint. It
also imports an assumption that does not hold: the design principles developed
for common-pool resources are framed for goods that are subtractable, so that
one party’s use diminishes what is available to others
(Ostrom, 1990), and a capacity may not be subtractable at all, as
Problem ? records.
Ownership by the configuration. The configuration holds the interest
itself. This requires that the configuration be a rights-holder, which requires
constitution as a legal person, which adds decision procedures, organs, and an
entry and exit regime. The entity that then holds the interest is not the
configuration that generated it.
Proportional shares. Each party holds a share. This requires a
decomposition of the jointly generated value, and the difficulty is old: in
team production the joint output exceeds the sum of separable contributions and
the marginal product of each input is not separately observable
(Alchian, 1972). Definition ? makes the stronger
claim that no decomposition exists to be observed.
Stewardship. A party holds the interest for the configuration. This
requires selecting the steward, and the authority to select is the authority in
question. The rule is circular at the point where it does its work.
The problem is whether these are four failures or one. The author’s inclination,
stated as an inclination, is that they are one. Each rule requires a
determinate answer to the question of who holds how much of what, and each
fails at the point where that answer would be supplied. The rule of unallocated holding answers that nobody holds any of it, and
thereby denies the question in place of answering it. Configuration ownership answers “the configuration holds all of
it” and has to constitute an entity capable of holding, which changes what is
held. Proportional shares answers directly and requires the decomposition
denied. Stewardship defers the answer to a selection procedure that presupposes
it.
If that is right, the four horns are one horn seen from four sides, and the
horn is that allocation presupposes decomposition.
Consequences.
The reduction, if it holds, converts a catalogue into a theorem and identifies
where a solution would have to intervene. It also supplies the test named at
Section 4: any proposed allocation rule requires
examination for
where it supplies the decomposition, and the examination is not satisfied by a
rule that supplies it implicitly.
The literature has not made this reduction. The argument from team production
treats non-separability as a problem of metering and of information cost,
dissolved in principle by a monitor (Alchian, 1972). The present claim is that
the decomposition is unavailable rather than expensive.
Defeater.
Problem ? is the most serious candidate and is stated
separately. Failing that, an allocation rule that is not one of the four, or a
demonstration that one of the four does not require what is claimed. The author
notes that the metering-cost reading of team production, if correct as against
Definition ?, would defeat this entry: on that reading the
decomposition exists and is merely unobserved, which is a different situation
entirely.
Can distinct parties hold distinct incidents in respect of a single capacity,
so that allocation proceeds without decomposition?
Derivation.
Ownership decomposes into incidents that may be held separately
(‘e, 1961), and jural relations are relations among persons
rather than between a person and a thing (Hohfeld, 1913; Hohfeld, 1917).
Nothing in either analysis requires that a single party hold all the incidents.
The divisibility of incidents suggests an escape from
Problem ?. One party might hold control over the
conditions of exercise, another a provenance interest,
another a right of access, another an immunity against destruction, without any
party holding a share of the capacity and without the capacity being divided.
The value is not decomposed; the incidents are, and the incidents are not
portions of the value.
The proposal is serious and the author cannot dismiss it. It is the strongest
objection to Problem ? available and has not been tested.
Two difficulties are visible without settling it. Each incident requires a
determinate object: a right of access is access to something, an immunity is
immunity in respect of something, and the specification of that something may
reintroduce the decomposition at one remove. And the incidents may not be
jointly exhaustive here as they are for a thing: a party holding all of control,
access, and provenance may still not hold what the framework calls the capacity,
in which case the bundle picture describes the arrangements around the object
without reaching the object.
Consequences.
If this succeeds, Problem ? is closed and a substantial
part of the framework’s negative programme is relieved. Much of the legal
apparatus already operates this way, and a regime of divided incidents would be
constructible from existing materials rather than requiring new ones.
Defeater.
This entry is a candidate solution rather than an obstacle, so the question is
whether it works. A worked scheme of incidents for a generative capacity,
showing for each incident what its object is and how the object is specified
without decomposition, would close both this and
Problem ?. The author regards constructing that scheme as
the single most productive piece of work available on this list.
Can it be established that one holder would generate more value than another,
where the values generated are of different kinds?
Derivation.
A consequentialist argument for reallocating a capacity compares what one
holder would generate with what another would. The comparison requires a
measure common to both.
Where the generated values are of one kind the requirement is met. Where they
are of different kinds it may not be. A configuration may generate knowledge,
income, standing, care, coordination, and the conditions of further generation,
and there is a developed position that goods of different kinds admit different
modes of valuation and that the imposition of a single measure is itself a
substantive move rather than a neutral one (Satz, 2010; Radin, 1987).
The entry draws its importance from the narrowness of its requirements. It requires no premise
about uncertainty, no claim about the future, and no position on any problem
above it. Even granting perfect prediction of what each holder would generate,
the comparison does not go through if the generated values are not
commensurable. The consequentialist argument therefore fails at a point earlier
than the epistemic one, and fails for a reason that better information cannot
relieve.
Consequences.
This is the second independent branch. It also corrects a defect in the
author’s earlier reasoning. Confronted with a proposal to reallocate a capacity
toward higher expected future generation, he resisted it on grounds of
unpredictability, which is the weakest available ground and is in principle
relievable by better method. The present entry was available and stronger, and
he did not use it.
Defeater.
A demonstration that the values a generative configuration produces are
commensurable, or that a defensible common measure exists for the comparison at
hand even if not in general. The author observes that markets supply such a
measure where they exist, and that the configurations most at issue here are
those for which no market exists.
Where neither the case for reallocation nor the case for the present
arrangement can be established, what follows?
Derivation.
By Problems ? and ?, no party can
establish what a configuration would generate under a different holder. The
consequentialist argument for reallocation therefore cannot be started.
The error. The author took that result to support leaving the capacity
where it was. It does not. The unavailability is symmetric: the same argument
that defeats “the other party would generate more” equally defeats “the
present party would generate more.” Awarding the benefit of the unavailability
to the incumbent requires a further premise, that present holding is itself a
ground of entitlement, and the framework denies that premise at Settled
Result ?. Having held that origin confers no claim, the author
cannot hold that possession does. The inference was unwarranted and is recorded
here rather than silently corrected.
What remains. One response is available and it is procedural rather than
substantive. Where a question cannot be resolved on its merits, a legal system
allocates the burden of proof and decides against the party bearing it. That
does not answer who should hold the capacity. It answers who loses when nobody
can show anything, which is a different and more modest thing.
The residual problem is that the allocation of the burden is itself a
substantive choice with distributive consequences, and the framework supplies
no ground for making it. Placing the burden on the party seeking change
reproduces the incumbent bias by another route. Placing it on the holder would
require a justification the framework equally lacks.
Consequences.
If no ground for allocating the burden is available, the framework’s contribution
in this region is confined to showing that the substantive question has no
answer, without indicating what a decision procedure should do.
Defeater.
A principled basis for allocating the burden of proof that does not covertly
reinstate either possession or origin as a ground of entitlement. The author
notes that existing law allocates such burdens on grounds of access to
evidence, and that access to evidence about a configuration is ordinarily held
by the party controlling it. That observation points toward placing the burden
on the controlling party, and the author records it as a direction rather than
a conclusion, since it has not been checked against the framework’s
commitments.
8. Layer 4: Attribution of Loss
This section states four entries concerning remedy. Each presupposes that the
allocation questions of Section 10 have determinate objects,
which that section leaves open; the assumption is granted here because the
difficulties recorded below are additional to it. The method is to take the
elements a claim for remedy requires and to ask what supplies each. Attribution
is treated at Problem ?, the head of damage at
Problem ?, the level at which legal certainty operates at
Problem ?, which records a defeated repair, and the
underlying legal interest at Problem ?.
Does counterfactual attribution remain available where a capacity and the
conditions of its exercise co-evolve?
Derivation.
Attribution of loss proceeds counterfactually. It asks whether the outcome
would have obtained had the defendant not acted, and it obtains an answer by
removing one factor while holding the remainder fixed.
The operation requires that the factor be separable from the remainder. Where
the conduct complained of has itself shaped the conditions under which the
capacity developed, and the capacity has in turn shaped those conditions, the
factor cannot be removed while the remainder stands. What removal would take
away is partly constitutive of what would remain, and the subtraction has no
determinate result.
The point is worth distinguishing from a familiar one. It is often said that
causal questions of this kind are difficult, that they involve many variables
and long chains, and that judgement is required. That is a claim about
tractability. The present claim is different: the operation that attribution
performs requires a separation which, under Equation (2), is
not present. The limitation lies in the instrument rather than in the reasoner: attribution
requires an independence that the object withholds.
Consequences.
Any remedy predicated on showing that a defendant’s conduct diminished a
capacity inherits this. Problem ? states a form of
damage that does not require showing what would have been generated, and it
does not escape the present problem, since establishing that a contraction
occurred still requires distinguishing the contraction from what would have
happened anyway.
Defeater.
Whatever defeats Problem ? defeats this. Separately, a
form of attribution that does not proceed by removal would suffice. Doctrines
that shift a burden where the defendant’s conduct has itself made proof
impossible have this character in outline, and whether they extend to a case
where the impossibility is structural rather than evidential is a question the
author has not resolved.
Is there a form of damage that consists in the contraction of a capacity to
generate, as distinct from the loss of any particular thing generated?
Derivation.
Suppose a party’s space of available future relations contracts, as at
Equation (3), without any determinate outcome being prevented.
$$\opt’ \subsetneq \opt$$
Nothing specific has been lost. What has been diminished is the system that
would have produced things not yet specified.
The existing heads of damage accommodate a loss of a different shape. Lost profit requires an
identified stream. Expectation damages require a promised performance. Loss of
a chance, where it is available, requires an identified opportunity whose
probability can be stated. Each presupposes a specified object whose
non-occurrence is the loss. Here the object is unspecified by construction, per
Problem ?, and specifying it would convert the generative
capacity into a production forecast.
The question is whether a head of damage can be constructed whose object is a
contraction of range. Two features would distinguish it. Its measure would be
structural rather than monetary in the first instance, since what contracted is
a set of available paths. And its proof would concern the alteration of
conditions rather than the non-occurrence of outcomes.
Consequences.
This is the point at which the framework’s descriptive results would acquire
remedial consequence, and its absence is why they presently do not. The
companion work identifies arrangements in which a configuration’s capacity to
generate is consumed while every transaction passes inspection. Without a head
of damage of this kind, that identification supports no claim.
Defeater.
A worked head of damage, or a demonstration that an existing one extends.
Injunctive relief is the more promising route and is discussed at
Problem ?, since an obligation not to contract a range may
be statable where a monetary measure of the contraction is not. The author
regards the construction of such a head as the most valuable practical work
available on this list, alongside the scheme of incidents at
Problem ?.
Does the unavailability of counterfactual products leave claims about a
grammar’s persistence available?
The repair as proposed.
Legal certainty requirements operate on outcomes. A claimant is required to
show what would have been earned, produced, or received, and
Problem ? shows that such a showing is unavailable for
generative capacity.
The author proposed that this measures at the wrong level. A grammar is more
stable than its instances; that is what makes it a grammar. Any particular
sentence of a language is unpredictable while the language remains the same
language from year to year, and the two facts are consistent. If so, the
following might be provable without predicting any product: that the
configuration previously generated in a certain manner, which is a fact about
the past; that an intervention removed a condition on which that manner of
generating depended, which is a claim about structure; and that generation of
that manner has since ceased, which is a further fact. Together these appear to
establish destruction of a capacity with no counterfactual product anywhere in
the argument.
Failure of the repair.
It requires that the manner of generating be identified, and by
Problem ? identification is harder than prediction rather than
easier. A grammar observed only under one set of conditions is compatible with
many grammars that differ elsewhere. When conditions change and generation of
the observed kind ceases, three descriptions fit the evidence equally: the
capacity was destroyed; the capacity is intact and the new conditions do not
instantiate it; the capacity has become a different one whose products are not
recognized as of the same kind. Nothing distinguishes them.
The repair therefore relocates the difficulty without relieving it. Worse, if
Problem ? has no affirmative answer, the three-way
distinction is malformed rather than merely undecidable, since it presupposes
that capacity and conditions can be separately identified.
Grounds for retention.
Two reasons. The failed repair locates the difficulty more precisely than the
original statement did, by showing that moving to the level of the grammar does
not help and why. And the repair contains a correct observation that survives
its failure: the division between the rule level and the case level in existing
law is principled rather than inconsistent. A legislator may hold that a class
of incentive increases a class of generation, which is a claim about a
distribution; a court may not hold that this party would have generated this
thing, which is a claim about an individual counterfactual. Generativity is
unavailable at the second and not necessarily at the first.
Remaining doctrinal routes.
Loss of a chance, in jurisdictions and contexts where it is available, awards
for the loss of an opportunity without requiring proof that the outcome would
have obtained. Whether it extends to the loss of a capacity to generate
unspecified opportunities is open, and this route is the most likely doctrinal
approach to Problem ?.
Goodwill valuation shows that courts accept a price for a capitalized
expectation of future generation. The acceptance is qualified in a way that
matters here: what is accepted is a valuation the market has made, not one the
court has made. Goodwill is also treated as inseparable from the business to
which it attaches (Lords, 1901). Configurations with no market, which
include most of those at issue here, fall outside the qualification.
Injunctive relief is forward-looking and requires no monetary measure. An
obligation not to destroy may therefore be statable where a claim for the value
of what was destroyed is not. This is the most promising of the three and
depends entirely on Problem ?, since an obligation not to
destroy has no content without a criterion distinguishing destruction from
development.
What legal interest, if any, is constituted by an accumulated generative
relation?
Derivation.
The generative artificial intelligence case makes the difficulty enumerable,
which is why it is used.
Four things are stored and can be listed. The model is stored, as parameters.
The interaction history is stored, as text, which a user may typically read,
export, alter, or delete without any parameter changing. The outputs are
stored. The infrastructure that brings model and history together is
implemented in code.
The capacity that a long-standing user and a model jointly constitute is not on
that list. It is not a stored object at all. It is a disposition that exists
when the model and the history are brought together in an act of generation,
and between acts it is nowhere. The enumeration is what makes this more than an
assertion: in other settings the claim that a capacity resides in a
configuration rather than in a place is a metaphysical thesis requiring
argument, and here the components can be listed and it is not among them.
The legal position around this absence is instructive. Existing categories
attach to the enumerable items. Outputs raise questions of authorship, and in
the United States the human authorship requirement means that material
generated without sufficient human contribution is not copyrightable at all,
with registration guidance requiring that such material be disclaimed
(Circuit, 2025; Office, 2025; Office, 2023);
other systems reach different results by other routes: the United Kingdom deems the author of a
computer-generated work to be the person who made the arrangements necessary
for its creation (Kingdom, 1988), and a Chinese court has found an
AI-generated image copyrightable on the basis of the human prompter’s selections
(Court, 2023). Interaction history raises questions of personal data and of
portability. The model raises questions of trade secret and of contract.
The accumulated capacity raises none of these, because it is none of these
things. It is not a work, since it is not a determinate expression. It is not
personal data, since it is not a record about a person. It is not a thing to
which a right in rem could attach, for the reasons at
Problem ?. And a contractual allocation of the outputs
does not reach it: the providers’ assignment of output rights, qualified in the
operative terms as an assignment of such rights “if any”
(Henderson, 2025), allocates the products while leaving the capacity
and every condition of its exercise where they were.
The natural experiment.
The setting supplies a test that other settings do not. Export the interaction
history and instantiate it against a different model. The history transfers
completely; there is no loss in the copying, no ambiguity about what moved, and
none of the ordinary impediments to transferring practice within organizations
(Szulanski, 1996), since the transfer is mechanical. What follows is
a different capacity. Everything transferable was transferred and the result
differs, which is the signature of an object that is not of a transferable
kind.
The test is cheap and repeatable, which distinguishes it from the acquisition
case, where a failure of capability transfer can always be attributed to poor
integration.
Consequences.
Whatever obligations exist in respect of an accumulated generative relation
have no object under existing categories. This bears directly on
Problem ?: a head of damage requires an interest, and
the interest is missing before the measure is.
Defeater.
A characterization of the interest that does not reduce to one of the four
enumerable items. The author notes that portability regimes give a right over
the history and expressly give nothing over what the history generates, and
that this is not an oversight: the right is given where an object can be
individuated and withheld where it cannot. That pattern, together with the
inseparability of goodwill from its business (Lords, 1901) and the
protection of trade secrets through relations of confidence rather than through
a right in a thing (States, 1917), suggests that law has met this
difficulty before and has responded the same way each time, which is a fourth
instance rather than a solution.
9. Layer 5: Obligations in Respect of the Object
This section states the three remaining entries. Problems ?
and ? are normative and stand independent of every preceding
entry: each arises whatever answer Problem ? receives, and
each would remain open if every other entry were closed.
Problem ? states a settled result and appears here for the
reason given in Section 4. The method for the two open
entries is to identify structures in which obligations run without ownership,
and to ask what their application here would require.
Can the destruction of something that cannot be owned be a wrong, and if so on
what ground?
Derivation.
Suppose the negative results hold: a generative capacity is not an object of
property, no allocation of it can be made, and no party holds it. It does not
follow that anything may be done to it. The inference from unowned to freely destructible requires the premise that
wrongs against a thing are wrongs against its owner, and the standing of that
premise remains open.
Structures exist in which obligations run without ownership. A fiduciary owes
duties in respect of interests the fiduciary does not own, and the duty is
owed to a beneficiary rather than derived from the fiduciary’s holding.
Obligations of confidence protect a relationship rather than a thing; the
classical statement is that the property may be denied while the confidence
cannot (States, 1917). Environmental obligations attach to objects no party
owns and are frequently owed to no determinate party at all.
The structure is therefore available. What is missing is its application here,
and two specific gaps prevent it.
The first is the identification of the party to whom the duty would be owed. A
fiduciary duty runs to a beneficiary and a confidence to a confider. If the
capacity belongs to a configuration, and the configuration is not a person,
then the duty has no obligee unless one is constituted, and constituting one
raises Problem ? in its second form.
The second is the content of the duty. An obligation not to destroy requires
the distinction at Problem ?, and an obligation to preserve
raises Problem ?.
Consequences.
Without this, the framework’s descriptive results have no normative
consequence whatever. It can say that an arrangement consumes a capacity while
every transaction passes inspection, and it cannot say that anything has gone
wrong.
Defeater.
An account of duties in respect of unowned generative capacity, specifying the
obligee and the content. The author observes that the environmental case is the
most promising analogue precisely because it dispenses with a determinate
obligee, and that it does so by statute rather than by derivation from a
general principle, which may be the honest lesson.
Is any party under an obligation to maintain a generative capacity, as distinct
from an obligation not to destroy one?
Derivation.
This is the furthest entry on the list and the author has least to say about
it.
If generative capacity has value, and if no party owns it, the question arises
whether anyone is required to sustain it. The question is not answered by
Problem ?: a duty not to destroy is negative and requires
only forbearance, whereas a duty to maintain is positive and requires
expenditure, and the grounds for the two differ in the ordinary case.
A difficulty specific to this entry should be recorded, since it may show that
the duty has no content even if it exists. Under Problem ?
a capacity is not stored anywhere; it is a disposition obtaining when
components are brought together. Altering any component may therefore alter it.
The reasons for altering components are ordinarily good ones: correcting
defects, improving performance, reducing risk, responding to circumstances that
have changed. If almost every alteration has such a reason, then an obligation
qualified as one not to destroy without cause may be satisfied
everywhere and bind nowhere.
The author does not regard this as settled. It may be that the qualification is
the wrong one, and that what is owed is notice, or an opportunity to migrate,
or a preservation of the components rather than of the capacity. Each of these
has content that survives the difficulty, and each would need to be argued for
separately.
Consequences.
This entry bears on what a continuity interest could amount to in the
generative artificial intelligence case and in every case with the same
structure. If it has no answer, then the most that can be claimed there is
portability of the enumerable components, and the capacity itself supports no
claim of any kind.
Defeater.
Either a positive account of a duty to sustain, or a demonstration that the
obligations available in this region are exhausted by notice, portability, and
forbearance, so that the question does not arise. The author regards the second
as the more likely outcome and states it as an expectation rather than a
conclusion.
Does contributing to the constitution of a party or a configuration ground a
claim over what that party or configuration subsequently generates?
Status.
This entry is not open. It appears in the numbered sequence for the reason
given in Section 4. A document recording only difficulties
gives a reader no way to calibrate its other entries, and this result is
repeatedly under pressure from arguments that would otherwise be admitted
elsewhere in the list.
The result.
Origin confers no claim. That a party brought a configuration into being, or
contributed to the constitution of a party who subsequently enters other
configurations, grounds no claim over what those configurations generate.
The argument.
A but-for account of entitlement does not terminate. Every configuration has
constitutive ancestry without limit: those who constituted the parties, those
who constituted them, the language in which every subsequent interaction is
conducted, the institutions supplying the conditions. If contribution to
constitution grounded a claim over subsequent generation, every party would
hold claims over every other, and what would determine who collected is who
was positioned to press a claim. That is a theory of collection rather than of
entitlement.
A second argument runs from Definition ?. A claim to a share
requires a decomposition of jointly generated value, and the constituting party
was not a member of the configuration that generated it. The founding
commitment denies that its contribution can be decomposed out.
A third is structural. The constitutive contribution went into a party. The
value at issue was generated by a configuration of which that party is one
member. Standing follows participation in the generating configuration rather
than causal ancestry of one of its members.
Grounds for defence.
The result is symmetric and its symmetry is what makes it stable. It denies
that a constituting party holds claims over what the constituted party
subsequently generates. It equally denies that a party who has contributed over
time to the accumulation of a capacity thereby holds a claim over the
capacity. Applied in one direction only it would be a convenient position
rather than a principle, and the pressure to apply it in one direction only is
considerable, since the two directions favour different parties in the cases
that matter.
Defeater.
An account of a claim arising from constitution that terminates, and that does
not require the decomposition denied at Definition ?. The
author notes that a claim to reciprocity within the constituting relation is
untouched by this result and is not denied: what is denied is the conversion of
such a claim into an interest in a different configuration.
10. Settled Results of the Framework
This section states the four conclusions the framework establishes, together
with an account of the division between its settled and its open results. The
section serves a calibrating function: a list of open problems standing alone
would misrepresent the position of the framework, suggesting that it has
determined nothing, and a party favoured by an existing arrangement would find
that suggestion useful. Each result appears with its argument compressed; the
companion papers supply the full versions. The section closes by identifying
the appropriate targets for a reader who holds the framework to be mistaken.
Four results are settled in the sense that the author regards the arguments as
sound and is not aware of a defeater. They are stated with the argument
compressed; the companion papers supply the full versions.
10.1 Statement of the results
Any assignment of shares in a jointly generated capacity presupposes a
decomposition of that capacity into contributions attributable to parties who
existed beforehand. Definition ? denies that such a
decomposition exists. The consequence is negative and determinate: proportional
allocation is unavailable, and the unavailability is a matter of principle
rather than of measurement difficulty.
The qualification at Problem ? should be kept in view.
What is settled is that proportional allocation fails. Whether the other three
rules fail for the same reason is the open part, and Problem ?
may show that allocation without decomposition is possible by another route.
Holding rights in what a configuration has produced entails nothing about
rights in its capacity to produce further. The two are distinct objects, and no
argument carries from one to the other.
The generative artificial intelligence case makes this concrete. A user may
hold the outputs, by contract and subject to whatever the law recognizes, and
hold nothing whatever in respect of the conditions under which further outputs
may be generated. The reverse also holds: a provider’s ownership of a model
entails nothing about the capacity that model and a particular history jointly
constitute.
Stated and argued at Problem ?. Contribution to the constitution
of a party or configuration grounds no claim over what that party or
configuration subsequently generates, and the result applies in both
directions.
Where a position determines the forms in which value may be recognized, what a
configuration generates outside those forms receives no low valuation. It
receives no term, and a claim resting on it has no place to be lodged. The
companion work develops this; it is settled and is the basis of several
descriptive results.
10.2 Division between settled and open results
The division has a shape worth naming. The framework settles negative results and leaves positive ones open. It
establishes the unavailability of certain allocations, the failure of certain
inferences to carry, and the absence of a vocabulary for certain losses. It leaves open what should be put in
their place. The division follows from the framework’s commitments and reflects
something other than the distance the work has travelled. A framework whose
founding commitment is a denial of decomposability will generate negative
results readily, since most positive proposals in this region require the
decomposition. The positive results, if they come, will come from
Problem ? or from Problem ?, both of
which look for structures that do not require decomposition rather than for
better decompositions.
Settled Results ? through ? are the
appropriate targets for a reader who thinks the framework wrong. They carry
commitments, and refuting one of them costs the framework more than closing any
open problem would. A reader who points to the length of the list has
established nothing, since the list is offered and not concealed.
11. Antecedents
This section records the traditions in which several of the entries above have
prior statements, and identifies for each the boundary between what those
traditions establish and what the present paper adds. The paper claims priority
in none of them. The method is to take each entry with a known antecedent,
name the source, and state the difference where a difference exists. Two
qualifications arising from survey appear at the end of the relevant
paragraphs, since each weakens support the paper might otherwise be taken to
have.
Non-separability of jointly produced value.
Problem ? restates, for capacities, a difficulty long
known for outputs: in team production the joint product exceeds the sum of
separable contributions and no marginal product is separately observable
(Alchian, 1972). The difference is in the modal status claimed.
That literature treats non-separability as a problem of metering and
information cost, in principle relieved by monitoring.
Definition ? claims the decomposition does not exist. If the
metering reading is correct as against the framework, Settled
Result ? fails, and the author records this as the most
consequential unresolved disagreement between the framework and an established
position.
Non-extractability of organizational capability.
Problems ? and ? restate, in a legal
register, the claim that routines function as organizational memory and cannot
be extracted from the organization that enacts them
(Nelson, 1982), which rests on the prior observation that a
knower knows more than the knower can tell (Polanyi, 1966), together
with the finding that the transfer of practice is impeded even within a single
firm (Szulanski, 1996). The mechanism is theirs. What the present paper
adds is the proposal that this bears on what can be an object of property
rather than only on what can be managed, and the enumeration argument at
Problem ?, which converts a metaphysical thesis into a
checkable one in the one setting where the components can be listed.
Inalienability.
Problem ? sits beside a developed literature on why
certain entitlements resist transfer, including the taxonomy separating
non-transferability from non-salability (Radin, 1987), the treatment
of inalienability as a distinct mode of protecting an entitlement
(Calabresi, 1972), and the parameters marking markets as noxious
(Satz, 2010). The paper’s framing differs and the difference is worth
stating: that literature is predominantly normative, holding that certain
goods ought not be sold. The present entries are ontological, holding that the
individuation and severability conditions any property claim presupposes may
not be satisfied. A reader who prefers the normative framing may reconstruct
several entries in it, at the cost of inheriting the burden of showing what
ought not be sold and why.
The apparatus of ownership.
Problems ? and ? use the analysis of
ownership into separately holdable incidents (‘e, 1961) and
the reduction of property to relations among persons
(Hohfeld, 1913; Hohfeld, 1917), together with the account of
property rights as in rem and of the standardization that follows
(Merrill, 2000). These supply the test that
Problem ? applies; they are not modified here.
Non-severable relational value in existing law.
Problem ? observes a pattern rather than discovering it.
Goodwill is treated as inseparable from the business to which it attaches
(Lords, 1901), and a mark carries no significance apart from the goodwill
it symbolizes, so that an assignment without the business is void
(Circuit, 1984); trade secrets are protected through a relation of
confidence rather than by a right in a thing (States, 1917); personality
theory grounds interests that resist transfer in the connection between a
creator and a work (Hughes, 1988). The author notes one
qualification arising from survey: the rule against assignment of a mark in
gross rests substantially on protection against consumer deception rather than
on non-severability, so the trademark limb is weaker support than the goodwill
limb and is cited accordingly.
Capacity as a peculiar object of transaction.
The observation that a capacity transacts differently from a thing, being sold
for a period rather than conveyed, and remaining inseparable from the person
whose capacity it is, is stated in the analysis of labour power
(Marx, 1976). Problem ? generalizes the
structure from a person to a configuration.
Option sets and exit.
Problem ? uses a contraction of the space of available
relations as the object of a loss. Assessing advantage over the set of
achievable functionings rather than over holdings is established
(Sen, 1985), as is the analysis of exit cost and its effect on
the burden borne by voice (Hirschman, 1970), and the requirement that
authority over a party hold open what that party will later be able to exercise
(Feinberg, 1980). The present use is a restriction of these to relations
rather than functionings or rights.
Relational ontology.
Definition ? rests on treating transactions and relations as
primitive and parties as what relations produce
(Emirbayer, 1997). The companion papers concede this at length.
The generative artificial intelligence structure.
Problem ? relies on the observation that providers assign
output rights in terms qualified as an assignment of such rights “if any,”
while the law may recognize no underlying right to assign
(Henderson, 2025), against the background of the human authorship
requirement (Circuit, 2025; Office, 2025) and the divergent
approaches taken elsewhere (Kingdom, 1988; Court, 2023). That mismatch between
contractual allocation and absent underlying right is documented and is not
claimed here. What is added is the extension from outputs to the capacity, and
the observation that the same structure appears wherever products are allocated
while the conditions of further production are retained.
12. Limits, Intended Use, and Conclusion
This section states the limits of the document, its intended uses, a summary
table of the entries, and the conclusion. The limits appear in descending order
of severity. The intended uses are given as four, together with one use the
document is written to resist.
12.1 Limits of the document
The limits appear in descending order of severity.
The list is one author’s. It records the problems encountered by a particular
line of inquiry, and the completeness of such a list cannot be established. A
problem absent here may be absent because it does not arise or because it was
not met.
The dependency structure is argued and not proven. Claim ?
asserts that six entries are appearances of Problem ?. The
assertion rests on each of them presupposing the separation that problem
questions, which is shown case by case rather than in general. A reader who
finds an entry that survives an unfavourable answer to
Problem ? has refuted the structure without refuting the
entries.
Several entries would dissolve under a different starting point.
Definition ? is a commitment, not a result. A framework
admitting that jointly generated value decomposes in principle, however
difficult the decomposition in practice, would find
Problem ? and much of Section 10
unnecessary. Section 14 records that an established position
takes exactly this view.
The defeaters are uneven. Some name a specific result that would close an
entry. Others name a direction. An entry whose defeater is a direction is
weaker as a problem, since it has not been shown what would settle it, only
where one might look.
Nothing here is formalized. Several entries invite formal treatment,
particularly Problems ?, ?,
and ?, and the absence of formalization means that the
arguments have been checked for consistency in prose alone.
The generative artificial intelligence material carries more weight than a
single domain should. It is used because its components can be enumerated, and
the enumeration argument at Problem ? has no counterpart
elsewhere. If that setting is atypical in some respect the author has not
noticed, several entries are weaker than they appear.
The legal material is stated from a limited survey and is uneven across
jurisdictions. Positions are given where the survey established them and are otherwise
omitted, and the paper makes claims only about the jurisdictions it names.
12.2 Intended uses of the list
Four uses are intended and a fifth is not.
As a research map. Claim ? identifies where work is
worth doing. Problem ? is the trunk and dominates six
descendants. Problems ? and ? are the
constructive entries, where a solution would be built rather than discovered,
and the author regards them as the most valuable.
Problem ? is closed by stipulation and could be closed
tomorrow.
As a test on proposed solutions. Section 4 states
this use, which the author regards as the principal one. Any allocation rule offered for
this region is to be examined for where it supplies the decomposition that
Definition ? denies. A rule that appears to allocate without
decomposing has either solved Problem ? or concealed a
decomposition, and the list is the instrument for telling which.
As a constraint on the author’s own companion work. Each open entry
marks a place where those papers stop. Where a companion paper states a limit
that corresponds to an entry here, the correspondence is recorded in that
paper. A reader of the companion papers who finds a confident claim in a region
covered by an open problem here has found an inconsistency, and the author
would want to know of it.
As an invitation. A problem list invites solution more directly than an
argument does. Any entry closed is a better outcome than the entry standing.
Not as a demonstration that nothing can be decided. The framework
settles four results, stated at Section 13, and one entry
in the numbered sequence records an occasion on which the author drew precisely
that convenient conclusion and was wrong. Citing an entry as showing that a
question has no answer, where what it shows is that the author has not found
one, misuses the document.
12.3 Summary
Table 2 lists the twenty-one entries with their layer, their
dependence, and the character of what would close them.
| @p0.045p0.25p0.055p0.16Y@
| No. | Problem | Layer | Depends on | Character of a defeater |
|---|---|---|---|---|
| ? | Separability of capacity from conditions | 0 | Nothing | A weak-coupling approximation with stated interval and tolerance |
| ? | Persistence weaker than identity | 0 | ? | A continuity criterion supporting assertion without supporting ownership |
| ? | Identification of a grammar from products | 1 | ? | An identification result under stated restrictions |
| ? | The generative portion is unobservable | 1 | ? | A theory linking exercised properties to unexercised range |
| ? | Counterfactual generation | 1 | ? | Whatever defeats ?; or population-level comparison |
| ? | Evolution of a capacity | 1 | ? | Empirical regularities at class level |
| ? | Membership of the configuration | 2 | ? | A criterion not itself an exercise of the power it regulates |
| ? | Identity conditions of a grammar | 2 | ? | Identity conditions imported from dynamical systems |
| ? | Partial transferability | 2 | ? | Closing ?; or showing the transferable part is never generative |
| ? | The meanings of transfer | 2 | Nothing | Stipulation |
| ? | Transfer without depletion | 3 | Independent | Showing the generative component is rivalrous |
| ? | Allocation presupposes decomposition | 3 | ? | ?; or a fifth rule; or the metering reading |
| ? | Divided bundles as an escape | 3 | ? | A worked scheme of incidents |
| ? | Value across kinds | 3 | Independent | A defensible common measure for the comparison at hand |
| ? | Symmetric failure of proof (records an error) | 3 | ? | A principled basis for allocating the burden of proof |
| ? | Independence required for attribution | 4 | ? | Attribution not proceeding by removal |
| ? | Generative loss as a head of damage | 4 | ? | A worked head of damage; or extension of loss of a chance |
| ? | Measurement at the wrong level (defeated repair) | 4 | ? | Loss of a chance; goodwill valuation; injunctive relief |
| ? | The absent legal object | 4 | ? | A characterization not reducing to an enumerable item |
| ? | Wrongness without ownership | 5 | Independent | Duties specifying obligee and content |
| ? | A duty to sustain generation | 5 | ? | A positive account; or exhaustion by notice and forbearance |
| ? | Origin confers no claim (settled) | 5 | Independent | A terminating account of constitutive claim |
Table. Entries by layer, dependency, and character of defeater
12.4 Conclusion
A framework holding that value generated by a configuration has no
decomposition into prior shares will produce negative results in abundance and
positive ones with difficulty. This paper states where the difficulty begins
and what shape it has.
The shape is a trunk and two branches. One question asks whether a generative
capacity and the conditions of its exercise can be separately identified, and
six further problems are appearances of it if the answer is unfavourable. Two
problems stand apart and survive any answer: that a capacity conveyed without
diminishing its source defeats the exclusion and depletion on which the
property model rests, and that comparing what two holders would generate
requires a measure across kinds that may not exist. The normative problems
stand apart likewise.
Three entries are marked and the marks are the document’s calibration. One
preserves an inference the author drew and should not have, from the
unavailability of proof on both sides to the retention of the present holder.
One preserves a repair that was proposed and defeated, because the manner of
its defeat locates the difficulty better than its statement did. One records the
single result that is settled, and which requires symmetric defence, since
applied in one direction only it would be a convenience rather than a principle.
The claim advanced is that setting this out is a result. The framework’s
founding commitment is a denial, and a denial invites the reintroduction of
what it denies under other names. A list of problems stated in dependency
order, with the argument for each and the defeater for each, is the instrument
by which such a reintroduction becomes visible. A framework returning confident
answers in this region would be one that had smuggled the decomposition back
in, and the author includes his own error at
Problem ? because the smuggling is easiest to perform when
the conclusion is congenial.
Acknowledgments
The author thanks Ron Eglash for dialogue that helped inspire the wider inquiry
into generative justice and value circulation within which this work was
developed. Several entries below reached their present form through a process
in which a proposed resolution was advanced and then defeated, and the author
records that the defeats were as productive as the proposals. The problems,
arguments, structural claims, and errors remain the author’s responsibility.
References
Honoré, A. (1961). Ownership. Oxford Essays in Jurisprudence.
Hohfeld, W. (1913). Some Fundamental Legal Conceptions as Applied in Judicial Reasoning. Yale Law Journal.
Hohfeld, W. (1917). Fundamental Legal Conceptions as Applied in Judicial Reasoning. Yale Law Journal.
Merrill, T., Smith, H. (2000). Optimal Standardization in the Law of Property: The Numerus Clausus Principle. Yale Law Journal.
Radin, M. (1987). Market-Inalienability. Harvard Law Review.
Calabresi, G., Melamed, A. (1972). Property Rules, Liability Rules, and Inalienability: One View of the Cathedral. Harvard Law Review.
Satz, D. (2010). Why Some Things Should Not Be for Sale: The Moral Limits of Markets. Oxford University Press.
Alchian, A., Demsetz, H. (1972). Production, Information Costs, and Economic Organization. American Economic Review.
Nelson, R., Winter, S. (1982). An Evolutionary Theory of Economic Change. Harvard University Press.
Szulanski, G. (1996). Exploring Internal Stickiness: Impediments to the Transfer of Best Practice within the Firm. Strategic Management Journal.
Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.
Hughes, J. (1988). The Philosophy of Intellectual Property. Georgetown Law Journal.
Henderson, P., Lemley, M. (2025). The Mirage of Artificial Intelligence Terms of Use Restrictions. Indiana Law Journal.
Circuit, U. (2025). Thaler v. Perlmutter, 130 F.4th 1039.
Office, U. (2025). Copyright and Artificial Intelligence, Part 2: Copyrightability.
Office, U. (2023). Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence.
Kingdom, U. (1988). Copyright, Designs and Patents Act 1988.
Court, B. (2023). Li v. Liu.
Lords, H. (1901). Commissioners of Inland Revenue v Muller & Co’s Margarine Ltd.
States, S. (1917). E. I. du Pont de Nemours Powder Co. v. Masland.
Circuit, U. (1984). Marshak v. Green.
Marx, K. (1976). Capital: A Critique of Political Economy, Volume One. Penguin in association with New Left Review.
Hirschman, A. (1970). Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Harvard University Press.
Emirbayer, M. (1997). Manifesto for a Relational Sociology. American Journal of Sociology.
Sen, A. (1985). Commodities and Capabilities. North-Holland.
Feinberg, J. (1980). The Child’s Right to an Open Future. Whose Child? Children’s Rights, Parental Authority, and State Power.
Polanyi, M. (1966). The Tacit Dimension. Doubleday.