Jurisprudence and Justice in Generative Relational Systems - A Preliminary Position Paper on the Generation of Subjects, Causation, and Responsibility 【（Preliminary）Draft】

Posted on In , Views: Disqus: Word count in article: 5.8k Reading time ≈ 21 mins.

Abstract

Modern law possesses a refined apparatus for relations, structures, and causal chains, and criminal causation is among the most developed causal theories in any normative discipline. This paper argues that the difficulty presented by heterogeneous generative relational systems lies elsewhere, in the form that legal representation is able to take. Four claims are advanced. First, law generates the subjects it requires to be antecedent to the acts it attributes, and its capacity doctrines locate that generation in natural fact, official act, or individual condition, while a small family of status doctrines already locates it in relational pattern. Second, legal causal representation carries six requirements, namely apportionability, co-level control, reversibility, locality, rate, and a fixed grammar, and generative systems violate each of them independently. Third, those requirements are projected from the form of the available judgment: acyclicity follows from finality, a binary measure of contribution follows from the binary verdict, a fractional measure follows from divisible damages, and locality follows from territorial jurisdiction. Fourth, every attribution device in law is a control test, so a system that generates its objects at a level no participant occupies can be neither obligor nor obligee, and law reaches such systems only by fiction, by proxy insertion, by personification, or by channelling. A constructed case of several parties coupled through one generative system, with no communication among them, exhibits a configuration for which the doctrine of joint attribution has no category, since it supplies causal coupling in the absence of normative coupling. The paper closes on two positive questions, the recognition of relationally generated value and the conditions under which the generation of a relation can be known at all, together with the conditions under which its claims would fail.

Keywords: legal form; subject formation; attribution; causation; generative relational systems; legal personality; remedial form; epistemic opacity; relational value.

A note on the standing of this paper. This is a position paper, written in advance of a longer work, and its function is to state claims compactly enough that they can be attacked. It concedes at the outset that the difficulty it describes is one of representational form and carries no suggestion that legal reasoning about relations, structures, or causal sequences is unsophisticated. Where the argument turns on doctrine, the doctrine is drawn from Japanese, German, English, and international sources, and the citations are given at a level of generality appropriate to a position paper and require verification before any of them is relied upon. Section 11 states the conditions under which each principal claim would fail. Objection, correction, and counter-evidence are welcome at huangwanhong@serendip.ngo.

1. The Problem of Representational Form

This section states the problem the paper addresses, sets out the concession on which the argument depends, and describes the structure of what follows. The method throughout moves from a concrete process of relational breakdown upward to a mechanism, and from a mechanism to a principle.

The framework the paper extends may be given in its own terms. Generative Relational Being (GRB) is a theory of how subject, meaning, value, creation, and normativity co-emerge through generative relational processes. The present paper opens its jurisprudential branch, and the branch inherits the framework’s obligation to state its claims in a form that permits their revision.

The concession comes first, because the argument fails without it. Law is not naive about relations. Complicity, conspiracy, and organisational fault attribute liability across pluralities of actors (Hart & Honoré, 1985; Roxin, 2006); duties of care arise from relational position; 安全配慮義務 and Verkehrssicherungspflicht fix obligations by reference to the sphere a party controls; the doctrine of due diligence in international law governs conduct that a state does not itself perform. Nor is law naive about causal sequence. The theories of 相当因果関係 and 危険の現実化, Roxin’s objective attribution, and the common law’s treatment of intervening acts constitute one of the most developed causal apparatuses in any normative discipline, and they already ask whether a danger created by one party was realised through the acts of another. Law also holds objects that are neither events nor persons. Competition law protects a process and names its closure foreclosure; merger control prohibits a transition between structural states in advance of any harm; prudential regulation requires a bank to maintain a capacity to absorb loss; the safeguarding of intangible heritage takes the viability of a living transmission as its object.

The difficulty therefore is not an absence. It concerns the form that legal representation takes when the system in question generates its own subjects, its own causal pathways, and its own conditions of visibility. Four sites are treated in turn: the generation of legal subjects (§3), the requirements carried by legal causal representation (§4) and their derivation from the form of the available judgment (§5), and the conditions under which a duty can be imposed at all (§6). A constructed case follows (§7), then two questions that the sites above leave open, namely the knowability of generation (§8) and the recognition of relationally generated value (§9). Section 11 states the conditions of refutation and the open questions.

2. Relation to Prior Work

This section identifies the bodies of work that occupy the ground adjacent to the argument and states what each of them already settles, so that the residue claimed later is narrow. The order is from the most general to the most specific.

Relational legal theory is established. Nedelsky’s relational account of the legal self (Nedelsky, 2011), Norrie’s relational critique of responsibility (Norrie, 2000), and the relational tradition in feminist legal theory hold the position that legal subjects are constituted through relations. The claim advanced in §3 is accordingly not that subjects are relational.

Distributed and shared responsibility is likewise established. Young’s social connection model (Young, 2011) supplies a forward-looking and non-liability account of responsibility for structural processes; Kutz (Kutz, 2000) grounds accountability in participatory intentions and thereby detaches it from difference-making; Floridi (Floridi, 2016) distributes responsibility across a network on a faultless basis; Miller (Miller, 2007) supplies the most systematic scheme for allocating remedial responsibility among many candidates; and the shared-responsibility programme in international law has examined cumulative and cooperative wrongdoing for over a decade (Nollkaemper & Jacobs, 2013).

The reflexive tradition already holds that legal translation transforms what it takes up. Teubner’s regulatory trilemma and his account of legal irritants (Teubner, 1983; Teubner, 1998), Luhmann’s systems theory (Luhmann, 2004), and the responsive-law tradition occupy that ground, and Teubner’s proposals on the status of autonomous software agents (Teubner, 2018) anticipate the question of intermediate legal status. The generative account of the subject is held by Foucault’s genealogy (Foucault, 1977), Bourdieu’s genetic structuralism (Bourdieu, 1990), Simondon’s theory of individuation (Simondon, 2020), and Butler’s account of subjection (Butler, 1997); actor-network theory holds that an actor is itself a network effect (Latour, 2005). Epistemic limits are held by Hayek’s knowledge problem (Hayek, 1945), Fuller’s account of polycentric tasks (Fuller, 1978), Burrell’s taxonomy of algorithmic opacity (Burrell, 2016), and Fricker’s account of testimonial and hermeneutical injustice (Fricker, 2007). Relational property is held by Singer (Singer, 2000), and by Dagan and Heller (Dagan & Heller, 2001). Two further positions bear directly on what follows: the requirement that a claim in private law be bipolar, correlating one plaintiff with one defendant, is defended by Weinrib (Weinrib, 1995), and the treatment of risk imposition as a wrong in its own right is developed by Oberdiek (Oberdiek, 2017).

Three residues remain, and the paper claims only these. The first is the derivation of legal representational form from remedial form (§5). The second is the observation that every attribution device in law is a control test, together with what follows for systems that generate their objects at an unoccupied level (§6). The third is the identification of a configuration, causal coupling in the absence of normative coupling, for which the doctrine of joint attribution has no category (§7).

This section distinguishes two theses about relational subjects, states which of them GRB holds, and identifies where law already locates the generation of a subject-position. The argument proceeds by definition, then by contrast with the nearest prior account, then by examination of the capacity doctrines.

Definition 3.1. (Structural generation). A subject-position is structurally generated when the relational structure in which it is occupied is itself produced by a process, and the properties of that process determine which positions are available to be occupied. The thesis of structural generation is distinct from the thesis that a subject is constituted in relations, since the latter takes the constituting structure as given and the former asks how it came to be.

Actor-network theory holds the ontological half of this already. What it declines, as a matter of method, is any criterion that would distinguish one formation from another. The criterion is what GRB adds, and the criterion is exit.

Claim 3.2. (Positions without exit). Structural analysis becomes legally operative where a formation closes exit from a subject-position, so that the occupant can no longer be generated otherwise. Elsewhere it remains descriptive background with no legal consequence.

This also marks the distance from the nearest prior account. Young’s structural injustice (Young, 2011) concerns structures that produce unjust outcomes. Claim 3.2 concerns structures that produce positions from which exit is closed, which is a different object and is not reducible to the first. Bullying supplies the clearest instance. A label, once congealed, operates upon every subsequent interpretation of its bearer, so that benign later conduct is absorbed as further confirmation. What has been injured is the capacity of the community to interpret this person otherwise, and no single act of the many that produced the label carries that injury.

Law has a mature and graded doctrine of when a subject-position exists. 権利能力 and Rechtsfähigkeit, 行為能力, 責任能力, 意思能力, guardianship, incorporation, and the status offences all address it. Their generators, however, are of three kinds only.

Proposition 3.3. (Generators of capacity). In the capacity doctrines the generator of a subject-position is a natural fact, an official act, or the condition of the individual. None of the three is a relational structure.

A small family of doctrines departs from this, and it is the family the argument builds upon. Employee status is determined by the pattern of subordination and direction, and not by the label the parties adopt; de facto marital status arises from a pattern of shared life; apparent authority arises from the appearance a relation presents to a third party; and the recognition of states, of insurgency, and of belligerency turns on relational facts together with the acknowledgement of others. Each of these generates a subject-position from a pattern, determined after the fact. That is the form the argument requires, and law already possesses it.

One consequence should be stated plainly, together with its guard.

Proposition 3.4. (Antecedence as a constitutive fiction). Legal attribution presupposes a subject antecedent to the act attributed. Where law is among the generators of the structure that produces subject-positions, the act of attribution participates in generating the subject to which it attributes. Generative systems accordingly do not defeat the requirement of antecedence; they make visible that it was always a construction.

The guard belongs in the same paragraph as the proposition. The construction is protective. Treating the subject as antecedent is what prevents retroactive and arbitrary attribution, and it is an achievement of legal form whose protective function this argument preserves. Nothing here licenses the conclusion that no one is answerable. What follows is narrower: the cost of the achievement becomes visible at the generative limit, and the position of the occupant of a generated position does not exhaust that occupant’s answerability for what he does from it.

This section sets out the requirements that legal causal representation carries, pairs each with the doctrine that imposes it, and states the claim that generative systems violate them independently. The method is doctrinal inventory, and the claim is refutable by exhibiting a single counterexample of the kind named in §11.

Proposition 4.1. (Representational requirements). Legal causal representation carries six requirements: apportionability, co-level control, reversibility, locality, rate-insensitivity, and a fixed grammar over the period adjudicated. Each is imposed by an identifiable feature of legal procedure or remedy, and each is violated by a property that generative relational systems exhibit.

Table 1 pairs them. The apportionability requirement forbids cycles: where two elements each contribute to the other, a measure of contribution is undefined, and the difficulty concerns the definition of a measure before it concerns its taking. The co-level requirement demands that the bearer of a duty occupy the level at which the injury occurs, which is developed in §6. Reversibility is presupposed by cessation and restitution, which assume that removing a cause restores a state, and hysteresis defeats that assumption. Locality is presupposed by territorial jurisdiction and by the law of the place of the wrong. Rate-insensitivity follows from law’s representation of durations without representation of rates, and above all without representation of the relative rate of two competing processes, which is what an account of absorption requires. The fixed-grammar requirement is presupposed whenever a court applies one scheme of interpretation across the period it examines, and it is defeated where the scheme itself is what changed.

Table. Requirements carried by legal causal representation, and what defeats each.

Requirement Imposed by Defeated by
Apportionability finality of judgment circular causation
Co-level control the control condition emergence across levels
Reversibility cessation and restitution hysteresis
Locality territorial jurisdiction topological distribution
Rate-insensitivity duration without rate competing process rates
Fixed grammar one interpretive scheme change of the scheme

Claim 4.2. (Independence). The six requirements are violated independently, which is why each doctrinal advance addresses exactly one. The substitution of material contribution to risk for causation of harm in the mesothelioma line (Honoré, 1999) addresses apportionability and touches none of the others.

5. The Projection of Remedial Form onto Causal Form

This section derives the requirements of §4 from the form of the judgment available, and states the corollary that follows for continuing remedies. The argument is by four derivations, each pairing a feature of remedial or procedural form with the representational constraint it produces.

Proposition 5.1. (Projection of remedial form). The structure of legal causal representation is determined by the structure of the available judgment. Acyclicity follows from finality, since a cyclic causal finding has no terminating node and no judgment can close upon it. A binary measure of contribution follows from the binary verdict, which is why criminal law treats an intervening act as breaking or preserving a chain and admits no third value. A fractional measure follows from divisible damages, which is why the same question admits proportional answers in tort. Locality follows from territorial jurisdiction.

The proposition reframes the entire inquiry. On this account the constraints of legal causal form are not epistemic deficiencies; they are projections of remedial form. The question that follows is accordingly constructive.

Corollary 5.2. (Relaxation under continuing remedies). Where the remedy is continuing rather than terminal, the requirements weaken. Structural injunctions, supervisory jurisdiction, guarantees of non-repetition, monitorships, and the conciliation-first design of family jurisdiction are the sites at which a court represents a process without having to close upon a terminating node.

A third instance of the same projection deserves separate statement, because it bears directly on the case of §7.

Corollary 5.3. (Burden as determinant). Where the predicate that classifies an intermediary cannot be evaluated, the outcome is fixed by the allocation of the burden of proof. The substantive answer is then produced by a procedural device.

6. The Control Condition and the Routes of Imposition

This section identifies the condition common to law’s attribution devices, states what follows for systems that generate their objects at an unoccupied level, and enumerates the routes by which law nonetheless imposes duties on such systems. The method is again doctrinal inventory followed by a claim about exhaustiveness.

Proposition 6.1. (The control condition). Every attribution device in law is a control test. The guarantor position rests on control over the source of danger; the safety duty of an employer rests on dominion over the workplace; the identification doctrine rests on the directing mind; parent-company liability rests on control of the subsidiary; attribution in the law of state responsibility rests on effective control. Law accordingly attaches duties, and rights, to loci of control.

Proposition 6.2. (Systems without a locus). A generative relational system produces its objects, among them labels, norms, and structures, at a level that no participant occupies. Such a system can therefore be neither obligor nor obligee, and one condition accounts for both failures: a commons cannot be a party because no one is competent to bind it, and a bullying dynamic yields no defendant because no one controls the grammar that injures.

Two mature bodies of law confirm the diagnosis by conceding it. Competition law cannot reach parallel conduct that produces the effects of agreement in the absence of agreement, except through a narrow route requiring conditions rarely satisfied; and the doctrine that refuses to trace a causal chain through the free and deliberate act of a third party is a considered protection against an infinite regress of liability, which is why the gain of a social network, composed as it is of the voluntary acts of others, is not treated as transmission.

Proposition 6.3. (Routes of imposition). Law imposes a duty upon a generative relational system by one of four routes. By fiction, freezing the process and treating a subject it produced as antecedent. By proxy insertion, relocating the duty to an exogenous party and assigning it a control it does not hold. By personification, conferring legal personality upon the process. By channelling, severing liability from both control and causation and designating a party who answers.

Claim 6.4. (Asymmetry of the routes). The four routes are not symmetric between rights and duties. Personification serves entitlement and fails for liability, since a sanction requires assets, continuity of will, and an addressable locus, whereas holding a right requires only a representative competent to invoke it. Rivers have accordingly been personified, and no bullying dynamic ever has been.

Proxy insertion carries a testable consequence, since the proxy is chosen territorially while the gain it is asked to govern is topological.

Hypothesis 6.5. (Decoupling). Compliance measures improve, and unrecorded failure persists, in proportion to the excess of assigned control over the proxy’s actual purchase upon the structure. The hypothesis predicts decoupling for a school charged with an online pile-on, for a contact point charged with the fourth tier of a supply chain, and for a territorial state charged with a decentralised network, and predicts its absence where purchase is genuine, as with bank capital and facility emissions.

7. A Constructed Case of Coupling Through Shared Generative Media

This section presents a constructed case, states what existing doctrine does with it, and identifies the configuration for which doctrine has no category. The case is a construction, offered to isolate one configuration, and §11 states what would follow if the configuration turned out to be doctrinally accommodated.

Several parties of different nationalities interact with one generative system whose outputs depend upon an accumulated and shared context. Two of them, $A$ and $D$, act with intent and in communication with one another. A third, $E$, contributes material in the ordinary course of work, with no knowledge of their purpose, and that material improves the system’s capacity to produce the outcome they seek. A fourth, $F$, converses on unrelated matters, and whether those exchanges contributed is unknown. A fifth, $G$, acts with intent, has no communication with $A$ or $D$, and does not know of their contribution. Harm follows, and the pathway cannot be established.

The greater part of this is tractable. $A$ and $D$ fall under the doctrines of common purpose. $E$ and $F$ fall outside criminal liability for want of the mental element, which is a protection and not a defect. The operator of a system that acts beyond its instruction is addressed by asking whether the danger created was realised, and the foreseeability of a type of harm suffices where the manner need not be foreseen.

Two observations survive that treatment. The first concerns the classification of the system itself. The question whether an intermediary interrupts a causal chain has never been a finding about an instance; it is a classification of a class, assigned by law. An animal’s behaviour does not interrupt the keeper’s liability; an employee’s does not interrupt the employer’s; a child’s interrupts partially; an independent contractor’s interrupts. The inability of law with respect to a generative system is accordingly legislative and not cognitive, and by Corollary 5.3 the outcome in the interim is fixed by the burden. There is one refinement worth stating, since it survives the foreseeability answer.

Claim 7.1. (The bounded repertoire). The distinction between the type of a harm and its manner presupposes a bounded repertoire of behaviour, within which an unforeseen manner may be absorbed under a foreseen type. A generative system has no bounded repertoire, so the type of the risk created cannot be specified without specifying what generativity denies.

The second observation is the one the case was constructed to isolate.

Claim 7.2. (Coupling without normative coupling). Every device of joint attribution in law requires normative coupling among the parties, whether as common purpose, conspiracy, joint enterprise, or co-perpetration, and one-sided co-perpetration is largely rejected. A shared generative medium supplies causal coupling among parties who have no relation to one another. That configuration, causal coupling in the absence of normative coupling, has no category.

Proposition 7.3. (The absent operator). The doctrine of intervening causes yields two outputs, interruption and preservation. A shared generative medium performs a third operation upon separate causal chains, merging them. Law possesses no operator of merger, and this is the same absence as the absence of an operator of aggregation across heterogeneous contributions, which is itself a failure of commensurability.

Where the pathway cannot be established, law already has a technology for harm that is certain and attribution that is not, namely channelled liability with a collective fund, as in the nuclear liability conventions, the oil pollution regime, and the no-fault compensation schemes. Channelling is the only one of the four routes that does not require the classification above to be resolved. Its price should be stated with equal clarity: caps, the absence of fault, and the disappearance of the individual wrong.

8. Sources of Generative Opacity

This section identifies the conditions under which the generation of a relation can be known at all, distinguishes three sources of its opacity, and states what the sources have in common. The material here is compressed, and the longer work treats it at length.

Opacity arises from complexity, where the scale, non-linearity, and adaptation of a system defeat reconstruction although every party wishes to understand it. It arises from power over interpretation, where certain paths of understanding are foreclosed, so that the account a person offers of himself is received in advance within a frame that absorbs it. And it arises from legal protection, where the law itself shields the opacity it then cannot penetrate, as with trade secrecy over a model and the contest over access to platform data.

Claim 8.1. (The common form). The three sources present one legal problem, since in each the generative trajectory cannot enter the admissible record. The claim of limited reconstruction is therefore institutional and concerns the record law admits, and it is not a claim about the cognitive capacity of persons.

Law has built an object for the second source, and the history is the argument. The harm of coercive control existed, was invisible to a law that recognised assaults, was named in scholarship, and became an offence whose unit is a course of conduct (Stark, 2007) producing constriction of a person’s space for action. Where the object resists that form, law substitutes conciliation and discretion for adjudication, which is what the jurisdiction over family disputes does, and which Corollary 5.2 predicts.

One guard is required wherever a person’s estrangement from the conditions of his own formation is asserted. The criterion is structural and it is not substantive. Nothing here determines what a person ought to believe about himself. What is asked is whether his own description of himself can enter the common record and be revised there.

9. The Protection of Relationally Generated Value

This section turns from harm to value, states what the ownership model presupposes, identifies where law already recognises relationally generated value, and formulates the two narrow gaps that remain. It closes with the criterion that any protection of such value requires.

The ownership model presupposes an attributable holder, exclusivity, transferability, and quantifiability, and its historical function in securing a person against the state and against others is not in question here. Consider two parties who have shared a life, and the photographs, letters, and diaries in which that life is recorded. One destroys them in the course of a quarrel. The claim in property is awkward, since the question of title does not reach what was lost; the claim in mental suffering is awkward, since it converts into money a value that was not generated in money; and the claim in personality interests is awkward, since the interest injured was held by neither party alone.

Law nevertheless does recognise value of this kind. The division of property upon dissolution of a marriage allocates value that neither party holds, on the basis of relational contribution including domestic and caring labour; compensation for injury to an object of special attachment is available in narrow classes; heritable objects, intangible heritage, and the knowledge of a community are all protected as value that no single holder generated.

Claim 9.1. (Two gaps). Law recognises relationally generated value at the dissolution of the relation and in monetary conversion. Two gaps follow: the value is unprotected against destruction while the relation persists, and its recognition is available only in a form into which it converts. The case above falls in the first gap, and the same value would have been recognised had the parties separated.

Claim 9.2. (Criterion of protection). Relational generation does not itself warrant protection, since a formation that requires the silence and obedience of its members is also relationally generated. The criterion is the one stated in Claim 3.2 and its companions: whether the formation generates capacity, admits participation, remains open to revision, and preserves a counterweight to the consolidation of power within it.

10. The Criterion of Justice and Its Restraint

This section states the criterion of justice the foregoing supports, marks the restraint the criterion observes, and distinguishes it from two positions with which it would otherwise be confused. The criterion is stated structurally, which is to say that it ranges over the openness of the positions a formation generates and takes no view of what any occupant of a position should believe or want.

Claim 10.1. (The criterion of justice). A generative relational formation satisfies the criterion where the occupant of any position it generates retains the standing to enter his own account of himself into the common record and to have that account revised there. The criterion ranges over the openness of positions and leaves the content of any party’s beliefs and wants untouched.

Two confusions are worth forestalling. The first would read Claim 10.1 as an argument about false consciousness, and the restraint just stated is what forecloses that reading: the criterion never holds that an occupant is mistaken about his own situation, and asks only whether his account of it can be heard and corrected. The second would read the criterion as a licence for the diffusion of answerability, on the ground that a formation rather than a person stands accused. Claim 10.1 carries no such consequence. Individual answerability remains wherever an act is locatable, and the criterion addresses a further object, which is the formation that determines what positions are available to be occupied at all. The two obligations sit at different levels and neither displaces the other.

The criterion also fixes the standing of the material in §8 and §9. Where opacity is produced by the control of interpretation, what is lost is precisely the standing that Claim 10.1 requires, and the loss occurs without any event that a tribunal could take. Where relationally generated value is at issue, the criterion supplies the ground on which some formations warrant protection and others do not, and it does so without conferring protection upon a formation merely because it is relational in origin.

11. Conditions of Refutation and Open Questions

This section states the conditions under which each principal claim would fail, and lists the questions the paper leaves open. It is placed last because the claims must be available before their refutation conditions can be read.

Proposition 4.1 fails upon the exhibition of a doctrine that takes a cyclic or a level-crossing object as its primary legal object. Proposition 5.1 fails if a jurisdiction is shown in which the representation of causal contribution varies independently of the form of the judgment available. Proposition 6.1 fails upon the exhibition of an attribution device that is not a control test. Claim 7.1 fails upon the exhibition of a doctrine that fixes the type of a risk without presupposing a bounded repertoire. Claim 7.2 fails if the configuration it describes is accommodated by an existing category of joint attribution, and the doctrine of concurrent independent offenders is the most likely candidate. Hypothesis 6.5 is testable against records of enforcement and of audit, and is falsified by the absence of the predicted divergence.

Four questions are left open. The first concerns the point at which the closure of exit becomes determinable in a form a tribunal could apply, since the criterion of Claim 3.2 does the discriminating work throughout and has not been operationalised here. The second concerns the classification a generative system should receive, given that Proposition 6.3 shows the choice to be legislative and that each of the four routes distributes its costs differently, the burden of a tool presumption falling in particular upon those who use autonomous systems as a means of participation. The third concerns the price of anticipatory attribution: displacing responsibility to a point before the harm is law’s standing answer to unprovable causation, and it is also the route by which the criminalisation of preparatory conduct expands, which is a cost this framework must count rather than welcome. The fourth concerns the formal representation of the rate of generation against the rate of absorption, which the present formalism cannot hold, since it fixes the interpretive scheme it works within and the phenomenon of interest is the alteration of that scheme.

A final remark on the standing of the whole. The requirements catalogued here are constitutive achievements of legal form, and they protect against arbitrary attribution. The argument identifies a class of objects those achievements cannot take, and the identification of that class is not a charge of failure. Nothing in the argument supports the inference that a system’s generativity relieves its participants of answerability, and any use of it to that end would invert it.

References

Bourdieu, Pierre (1990). The Logic of Practice. Trans. Richard Nice. Stanford: Stanford University Press.

Butler, Judith (1997). The Psychic Life of Power: Theories in Subjection. Stanford: Stanford University Press.

Burrell, Jenna (2016). How the Machine ``Thinks’’: Understanding Opacity in Machine Learning Algorithms. Big Data & Society, 3(1), 1–12.

Dagan, Hanoch, and Michael A. Heller (2001). The Liberal Commons. Yale Law Journal, 110(4), 549–622.

Floridi, Luciano (2016). Faultless Responsibility: On the Nature and Allocation of Moral Responsibility for Distributed Moral Actions. Philosophical Transactions of the Royal Society A, 374(2083), 20160112.

Fricker, Miranda (2007). Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.

Fuller, Lon L., and Kenneth I. Winston (1978). The Forms and Limits of Adjudication. Harvard Law Review, 92(2), 353–409.

Foucault, Michel (1977). Discipline and Punish: The Birth of the Prison. Trans. Alan Sheridan. New York: Pantheon Books.

Hayek, F. A. (1945). The Use of Knowledge in Society. American Economic Review, 35(4), 519–530.

Hart, H. L. A., and Tony Honoré (1985). Causation in the Law. 2nd edn. Oxford: Clarendon Press.

Honoré, Tony (1999). Responsibility and Fault. Oxford: Hart Publishing.

Kutz, Christopher (2000). Complicity: Ethics and Law for a Collective Age. Cambridge: Cambridge University Press.

Latour, Bruno (2005). Reassembling the Social: An Introduction to Actor-Network-Theory. Oxford: Oxford University Press.

Luhmann, Niklas (2004). Law as a Social System. Trans. Klaus A. Ziegert. Oxford: Oxford University Press.

Miller, David (2007). National Responsibility and Global Justice. Oxford: Oxford University Press.

Nedelsky, Jennifer (2011). Law’s Relations: A Relational Theory of Self, Autonomy, and Law. New York: Oxford University Press.

Nollkaemper, André, and Dov Jacobs (2013). Shared Responsibility in International Law: A Conceptual Framework. Michigan Journal of International Law, 34(2), 359–438.

Norrie, Alan (2000). Punishment, Responsibility, and Justice: A Relational Critique. Oxford: Oxford University Press.

Oberdiek, John (2017). Imposing Risk: A Normative Framework. Oxford: Oxford University Press.

Roxin, Claus (2006). Strafrecht Allgemeiner Teil, Band I: Grundlagen. Der Aufbau der Verbrechenslehre. 4th edn. Munich: C. H. Beck. [5th edn 2020, with Luís Greco.]

Simondon, Gilbert (2020). Individuation in Light of Notions of Form and Information. Trans. Taylor Adkins. Minneapolis: University of Minnesota Press.

Singer, Joseph William (2000). Entitlement: The Paradoxes of Property. New Haven: Yale University Press.

Stark, Evan (2007). Coercive Control: How Men Entrap Women in Personal Life. New York: Oxford University Press.

Teubner, Gunther (1983). Substantive and Reflexive Elements in Modern Law. Law & Society Review, 17(2), 239–285.

Teubner, Gunther (1998). Legal Irritants: Good Faith in British Law or How Unifying Law Ends Up in New Divergences. Modern Law Review, 61(1), 11–32.

Teubner, Gunther (2018). Digitale Rechtssubjekte? Zum privatrechtlichen Status autonomer Softwareagenten / Digital Personhood? The Status of Autonomous Software Agents in Private Law. Trans. Jacob Watson. Ancilla Iuris, 2018, 35–78. [Longer German version: Archiv für civilistische Praxis 218 (2018), 155–205.]

Weinrib, Ernest J. (1995). The Idea of Private Law. Cambridge, MA: Harvard University Press.

Young, Iris Marion (2011). Responsibility for Justice. New York: Oxford University Press.