Question 1 // Does the root source-of-intent role require consciousness in its occupant?

what it asks

In ASK's method, a source of intent is the role that supplies or authorizes a system's purpose and the standard it runs on, and the root role is the source role at the head of a chain of authority. This question asks whether whoever holds the root role must be conscious, and in what sense. It does not ask whether a consequential system needs such a role, which ASK's published doctrine requires whether or not anyone is "home" in the system doing the work. The question began as ASK's own hypothesis, which the survey neither assumes nor rejects, carrying it as one candidate among rivals.

the main positions

The survey's main evidence on this question is three primary accounts, one of what an intention is and two of what makes a will one's own. J. David Velleman, in "What Happens When Someone Acts?" (1992), treats intentions formed outside present awareness as genuinely motivating. Harry Frankfurt, in "Freedom of the Will and the Concept of a Person" (1971), lets a will's desires be unconscious but makes the capacity for critical reflection on one's own will a condition of being a person. Christine Korsgaard, in "Self-Constitution in the Ethics of Plato and Kant" (1999), describes a rational will as conscious of itself as a cause. The survey's assessment is that, within these three formulations, dependence on consciousness differs by formulation, by sense and by level of analysis, so they establish no single requirement, though nothing shows that none exists.

Ned Block, in "On a Confusion about a Function of Consciousness" (1995), draws the distinction between phenomenal and access consciousness defined above, and David Chalmers counts deliberate control among the "easy" problems, apart from the hard problem. The survey rules neither reading. It notes that access may bear on how a role is exercised, which differs from authority to adopt the role's standard, and gives this material no weight in any recommendation until each claim is traced to its source.

Christian List, in "Group Agency and Artificial Intelligence" (2021), holds intentional agency, moral agency and phenomenal consciousness apart. He proposes consciousness, or the potential for it, as necessary for rights a being holds in its own right, without claiming it is necessary for fitness to be held responsible, a distinction the machine personhood survey rates the most precisely drawn in its sweep. Kenneth Einar Himma, in "Artificial Agency, Consciousness, and the Criteria for Moral Agency" (2009), holds that moral agency presupposes consciousness, and Luciano Floridi and J. W. Sanders, in "On the Morality of Artificial Agents" (2004), detach it from mental states, a contradiction the survey, reading both only in abstract, leaves unresolved. Patrick Butlin, Robert Long and colleagues, in "Consciousness in Artificial Intelligence" (2023), treat indicator properties as making machine consciousness more likely, not certain. Jonathan Birch, in "Large Language Models and the Gaming Problem", a chapter of The Edge of Sentience (2024), argues that behavioral tests fail once a system can mimic the signs of sentience.

David Deutsch, in The Beginning of Infinity (Penguin, 2011), defines people by their capacity to create new explanations and tentatively guesses that every genuine AI would be a person, with consciousness arriving together with that general capacity. In a 2023 dialogue with Deutsch on The Joe Walker Podcast, Steven Pinker disputes the step from computation to experience and expects the question to stay undecidable. The survey records the co-arrival claim as Deutsch's own conjecture and judges that a definition confers no phenomenal, moral, legal or institutional status. Keeping consciousness and personhood apart neither refutes nor confirms the conjecture, in the survey's view, since distinct things may still depend on one another. The survey does not adopt Pinker's expectation, and finds that neither speaker settles who may hold the root role.

what ASK ruled

ASK's disposition of August 23, 2026 re-held this question in a sharper form.

• Ruled: nothing.
• Kept as the survey's analysis, not as a ruling: occurrent conscious awareness (awareness of a particular attitude at the time) is not required by the formulations obtained that address it. Phenomenal consciousness is not established as required, and access consciousness in Block's sense is not established. A capacity-level reflexive relation to oneself is required by the obtained ownership and self-constitution formulations, at a separate point in each.
• Re-held: no mapping has been established from ASK's root source-of-intent role to those formulations.

what remains open

The question as a whole stays open. As the survey records, any argument linking consciousness to the root role must name its sense of consciousness. Primary texts by G. E. M. Anscombe, Donald Davidson, Michael Bratman and Gary Watson were not obtained, the two opposed accounts of moral agency were read only in abstract, and no source obtained is recorded as empirically judging whether any specific current AI system is conscious, has standing or could hold the source role.

Question 2 // Must chains of constituted authority ultimately terminate in natural persons?

what it asks

Authority is handed down chains, from a legislature to an agency or from a company to an automated system. The question is whether every such chain must finally rest on a natural person, a human being. ASK's published method doctrine deliberately holds it open, while requiring that answerability always stay locatable.

The survey separates the senses of resting on a person. Liability means being made to compensate or being convicted. Answerability, or accountability, is the duty to explain and justify oneself to a forum, which Cooper and colleagues' "Accountability in an Algorithmic Society" keeps distinct from compensating. Retributive responsibility is deserving punishment for having knowingly done wrong. The tracing condition, from Santoni de Sio and van den Hoven's "Meaningful Human Control over Autonomous Systems", requires that a system's behavior trace back to at least one person who understands its capabilities and recognizes accountability for its effects. Veil piercing is a court's exception to limited liability, setting aside a company's separate form so that a shareholder is held liable for the company's obligations.

the main positions

ASK put forward the hypothesis that answerability may need a human endpoint, reasoning from veil piercing. The survey's assessment is that piercing is a liability doctrine, so using it to reason about answerability crosses from one sense to another, and none of the reference works the survey read on veil piercing makes that move.

On liability itself, the issue is not whether a claim can reach through several corporate layers but whether every liability must end with a natural person. Identifying who owns or controls a company, through however many layers, does not by itself make the owner or controller liable. Piercing needs its own legal basis. As the survey records it, piercing engages only as a last resort, where a company is used to evade a legal obligation that already existed independently of it, and in the US only where the company is its owner's alter ego or instrument and there is improper conduct, against a strong presumption. The leading UK case, Prest v Petrodel Resources, is reported as declining to pierce.

The survey finds the hypothesis's real legal home in the identification doctrine, which, as the Law Commission of England and Wales records in its project on corporate criminal liability, convicts a company only through the acts of the senior people who represent its controlling mind and will, and so struggles with large, diffuse organizations. The Bribery Act 2010 convicts a company for failing to prevent bribery whether or not the briber is prosecuted, while the Act's separate route against senior officers covers other offenses and not this one. The survey reads this as making a human endpoint a design choice, offense by offense. The Corporate Manslaughter and Corporate Homicide Act 2007 bars individuals from being accessories, which the survey counts as decisive against necessity for that offense alone.

The EU's 2024 directive on defective products makes a chain of businesses liable without fault, and a distributor unable to name its supplier becomes liable itself. As the survey reads it, no natural person is needed, but no-fault liability is neither blame nor answerability.

John Dewey's "The Historic Background of Corporate Legal Personality" (1926) treats legal personhood as conferred by law. The survey records temples, a river and ecosystems granted personhood without being groups of people.

Santoni de Sio and Mecacci's "Four Responsibility Gaps with Artificial Intelligence" separates gaps in culpability, moral accountability, public accountability and active responsibility. In the survey's assessment, tracing supplies an anchor, a person who must understand but need not be punishable, and the EU AI Act's rule that natural persons oversee high-risk systems allocates oversight without saying how authority ends.

Christian List's "Group Agency and Artificial Intelligence" argues that organized groups are real agents and that humans answer through enacting, authorizing and design roles. Cary Coglianese's "Administrative Law in the Automated State" proposes a right to human empathy rather than to a human decision, which the survey treats as the strongest counterweight to assuming administrative authority ends in a human.

Pedro Domingos's The Master Algorithm (Basic Books, 2015) describes robot soldiers learning their ethics from examples screened by a panel of ethicists, with the manufacturer answering for malfunctions and the teachers for wrong calls. In its ruling of October 14, 2026, ASK made this account context only, with no weight on the question. The survey notes that naming answerers is not a complete remedy.

A rival hypothesis that the survey carried beside ASK's holds that answerability and recourse can run through institutions without reducing to one person who can be jailed. The survey finds it decisively supported on liability and compensation, and not established on answerability.

what ASK ruled

ASK's ruling of August 23, 2026, covers one strand.

• Ruled: Liability is one strand of "chains of constituted authority," and on that strand a chain need not end in a natural person. Veil piercing is retired as the mechanism of ASK's hypothesis on that strand.
• Left open: non-retributive answerability, punitive or retributive responsibility, and the tracing anchor, the condition that the system's behavior trace back to at least one person who understands it and recognizes accountability for its effects. ASK recorded that the liability regimes surveyed do not decide these.
• Expressly not ruled: this is not a ruling that chains of constituted authority need not end in natural persons. One strand is ruled, and the question is not.

what remains open

Everything beyond liability. The survey records that the ruling does not establish the counter-hypothesis, and that tracing and retribution, the two remaining candidates for a human endpoint, each need separate testing. The retributive question was never reached. The machine personhood survey covered English-language UK, EU and US material only, read no case law in the original, and read none of the central book-length works on collective agency, the machine question or legal personhood for AI.

Question 3 // May a computational office itself occupy the source role?

what it asks

This question asks whether an office run by, or consisting of, a computational system could itself hold the top authority-bearing role, rather than serve as a tool or a delegate under it. An office here is a constituted role: a position that exists because a governing order creates or recognizes it for a defined scope. ASK's published doctrine holds the question open. Its list of possible sources, which includes offices, and its ruling that the system doing the work may be human or computational do not together settle "whether a computational office may itself occupy the source role." The doctrine also holds that delegation never passes the source role downward.

the main positions

One line of evidence concerns legal personality, the status of being a person in law. Ugo Pagallo, in "Vital, Sophia, and Co." (2018), argues that legal agency and legal personhood are distinct, and notes that the European Union went almost two decades without legal personhood of its own. The EU's AI Regulation (2024/1689) supplies a second case, since its Article 3(47) defines the AI Office as a function of the European Commission, with no separate identity. The survey treats these as two direct counter-instances to the idea that holding institutional authority requires legal personality of one's own, while noting that the AI Office shows nothing about whether a non-human process could fill such a function.

Cary Coglianese, in "Administrative Law in the Automated State" (2021), argues that existing US administrative law can absorb extensive automation, meeting the duty to give reasons by explaining in general terms how an algorithm was designed and validated. In the survey's reading the machine stays an instrument throughout, with human agency supplied upstream by designers and overseers. Oliver Butler, in "Algorithmic Decision-Making, Delegation and the Modern Machinery of Government" (2025), faults doctrine for fixing on the final decision and recasts the matter as co-production between officials and private developers. As the survey reads him, he argues neither that a machine can exercise governmental power nor that it cannot.

Filippo Santoni de Sio and Jeroen van den Hoven's account of meaningful human control (2018) adds to the tracing condition described under Question 2 a requirement that the system respond to the relevant human moral reasons. The survey reads this as a design requirement that anchors control on a human being, an anchor rather than an end point, without attributing moral reasoning to the machine.

Ben Goertzel's research-program essay "Seeding RSI Toward ASI" (2026) lets a new mechanism gain operational control only after measured performance, keeps correctness and permission decisions with non-learned components in its test environments, bars learned components from arguing their way into wider permissions, and scopes the human role in high-consequence decisions to the period before AI reaches full human level. In the survey's assessment, within what was read, it establishes no basis for legitimacy, no office and no occupancy result in either direction.

In the dialogue with Pinker, Deutsch holds that artificial general intelligences will be people with rights, so that making them carry out our computations would be slavery, and that a system built to be creative cannot also be obedient. Pinker grants that such a system cannot be obedient about the problem set it, but sees no need for it to want autonomy in every respect. The survey's assessment is that recognizing personhood or welfare rights would not by itself confer command authority, and that, in Deutsch's book, neither creative choice nor the capacity to explain yields a result about who may hold the root role.

In the machine personhood survey, no source read makes any claim about non-natural occupancy of an office, and the works by Goertzel, Deutsch and Pinker supply no occupancy result either, so the central question is unaddressed.

what ASK ruled

ASK's disposition of August 23, 2026 re-held this question in a sharper form and settled one component of it.

• Ruled: an office does not need a separate legal personality of its own to hold constituted authority.
• Re-held: whether a computational or otherwise non-natural occupant may occupy the root source role remains open. The evidence addresses constituted offices and legal personality, not non-natural occupancy of the root role.
• Expressly not ruled: the partial ruling does not make this an answered question.

ASK's published doctrine records that the legal-personality result "does not establish root-role occupancy."

what remains open

The core ASK named stays open, and the survey records what it lacks. It read no source that addresses non-natural occupancy of an office, and read Ryan Calo and Danielle Citron's "The Automated Administrative State" only in abstract. Its free-will word searches could not detect authority expressed as standing, warrant, office or mandate, and it did not survey moral responsibility and the law, criminal-law theory or the political philosophy of desert.

The same doctrine holds open a neighboring question, whether chains of constituted authority must end in natural persons, treated separately on this page.

Question 4 // Is free will the same thing as intent?

what it asks

This was ASK's original question, and it turns on what "intent" means. In the psychological sense, intent is what a person means to do. In the governance sense ASK uses, the source of intent is a role carrying the standing to originate a directive that binds another actor or office. As the survey reads the literature, "free will" there names a metaphysical condition on moral responsibility, not the having of intentions, and its talk of an agent originating her actions concerns causal structure, not a governance role.

the main positions

Sourcehood accounts ask whether an action issued from the agent rather than merely through her. On the strongest reading, set out in the Stanford Encyclopedia of Philosophy's entry "Compatibilism" by McKenna and Coates, something necessary for the action must originate in her, not in earlier times and places. Kane, in "Libertarianism" (2009), makes free will the power to be the ultimate creator of one's own ends. The survey's assessment is that every sourcehood text it read frames sourcehood as a causal-historical relation between an agent and her own action.

Authority language does appear in this literature, at another level. The encyclopedia's entry "Free Will" by O'Connor and Franklin reports Lippert-Rasmussen's account of which attitudes are authorized to speak for an agent, and Bratman's planning theory gives self-governing policies a claim to speak for the person over time. The survey calls this authority intra-agential, because it settles which of my attitudes speaks for me, not who may bind others. The survey's word searches across nine works, which were not uniform and could not detect the idea when expressed as standing, warrant, office or mandate, found no use making institutional or role authority the condition of sourcehood. The survey treats this as a narrow result about the terms searched, not as silence across the field.

Pedro Domingos, in The Master Algorithm (Basic Books, 2015), argues that machines will not have independent wills. Engineered rather than evolved, they extend our will by maximizing an evaluation function we set, and surprise us only in pursuit of goals we give them. He hedges on whether humans have free will. The inference is his own. The survey's assessment is that control, motivation, locally generated goals and metaphysical sourcehood are separate questions, so engineering origin alone does not rule out a machine will.

Philosophy of action treats intent in the psychological sense. Mele, in "Free Will and Science" (2011), separates urges, decisions and intentions: deciding is forming an intention, and intending is more tightly tied to action than wanting. The survey draws the guard that a felt urge is not an intent, and records that it has not yet tested whether Anscombe's, Davidson's or Bratman's accounts of intention depend on consciousness.

Responsibility-practice accounts study the practice of holding people responsible. P. F. Strawson, in "Freedom and Resentment" (1962), argues that the metaphysical thesis is irrelevant to it, Scanlon, in "The Significance of Choice" (1986), declines to settle determinism through our convictions about responsibility, and Shoemaker and A. Smith do not raise the question. The survey records that none of them asserts that responsibility depends on free will, though reference works report that the field commonly assumes it, and it treats that silence as evidence in neither direction.

The survey also establishes that "could have done otherwise" is ambiguous between a conditional reading (she would have, had she chosen) and a categorical one (she could have, with the past and the laws of nature held fixed). It does not establish which reading is right, and Question 5 takes it up. Agent-causal libertarians such as Chisholm and O'Connor hold that the agent originates her free decisions, which the survey reads as a claim about causal structure, not a normative role.

what ASK ruled

ASK decided this question only for one stated sense of "intent". Where "intent" means the normative-governance source-of-intent role, free-will sourcehood is not identical to intent, and ASK classified the two as not the same thing. In ASK's terms both may organize questions of origination, agency and authorship attribution, but free-will sourcehood concerns the agent's relation to her own actions and motives, while source of intent is a governance role carrying standing to originate a directive that binds another actor or office. ASK left every other sense of "intent" unruled, psychological and action-theoretic senses included.

The research records that this classification carries no implication for Question 1 or Question 3, both still open. ASK's published doctrine adds that non-identity is not non-requirement: whether a separate free-will condition bears on eligibility for the root source role is not settled. The survey adds that its reading of the literature does not confirm the separation, since outside literature cannot silently confirm or revise ASK's own terms.

what remains open

Whether free will is intent in any sense other than the governance one stays open. Outside the ruled sense the research declines to assume the identity, and could answer the rest only by recording the argument, the rivals, the evidence and the remaining uncertainty.

Question 5 // Can a constrained will be free? Where does bounded discretion sit relative to the compatibilism debate — as an instance, an analogue, or neither?

what it asks

Compatibilists hold that free will can coexist with determinism, the thesis that every act follows from the laws of nature and the remote past. The question asks whether a constrained will can be free, and whether ASK's bounded discretion, judgment inside a delegated mandate, is an instance of metaphysical freedom as the compatibilism debate understands it (a genuine case), an analogue (a parallel with a stated break) or neither. "Constrained" can mean causal constraint, as in determinism, or authorization bounds set by a delegating authority. When the survey warns that bounded discretion is not unconstrained will, it uses "constrained" in the second sense only.

the main positions

Classical compatibilism begins with Hobbes and Hume, who treat freedom as acting on one's will without external impediment. As the survey reads it, that account supplies no alternatives, so the tradition added a separate conditional analysis: Moore, Hobart and Ayer read "could have done otherwise" as "would have, had she chosen", though Moore declined to treat that as enough for free will. The encyclopedia's "Compatibilism" entry reports that this analysis failed, since the conditional can hold of someone with no real alternative, which the survey reads as a failed analysis, not a refuted compatibilism. The new dispositionalists (Vihvelin, Fara, M. Smith) test ability with the agent's causal base held fixed.

On the categorical reading, van Inwagen's consequence argument, in "The Incompatibility of Free Will and Determinism" (1975), holds that if determinism is true, our acts follow from the laws and the remote past, which are not up to us, so our acts are not up to us. Lewis, in "Are We Free to Break the Laws?" (1981), replies that being able to do something such that a law would have been broken is not being able to break a law.

On the leeway axis, the Principle of Alternative Possibilities (PAP) makes moral responsibility require that one could have done otherwise. Frankfurt, in "Alternate Possibilities and Moral Responsibility" (1969), rejects it with cases where an intervener would force the choice but never acts, so the agent cannot do otherwise yet seems responsible, and the encyclopedia reports no clear victor among the replies. The survey records, as a live dispute, that leeway and sourcehood do not separate as cleanly as the standard taxonomy suggests, and that the PAP sources it read address responsibility for an act, not latitude inside an authorized mandate.

On the sourcehood and hierarchical-control axis, Frankfurt's "Freedom of the Will and the Concept of a Person" (1971) makes the will free when the desire that moves you is the one you want to move you, and Fischer and Ravizza require a reasons-responsive mechanism that the agent owns. Manipulation arguments from Pereboom and Mele, which hold that being determined is relevantly like being covertly manipulated, press every compatibilist account.

Ronald Dworkin, in "The Model of Rules" (1967), treats discretion as relative to a standard-setting authority and distinguishes two weak senses, in which the standards demand judgment or the decision is final, from a strong sense, in which the official is not bound by that authority's standards, which he denies is license. The survey records that the text does not address the will.

ASK's Governance doctrine defines delegated discretion, which the survey treats as the definition of bounded discretion, as bounded judgment in his first weak sense, where the standards demand judgment, by an authorized decision-maker within the governing standard and delegated scope. Bounded Generativity adds that weak discretion confers no finality or immunity from review, and that the judging decision-maker may be human or computational, though authorization alone does not suffice and current model sampling does not qualify.

Domingos holds that intelligence and will need not reside in one body so long as a line of control connects them, arguing from an analogy with genes and the organisms they shape that he himself qualifies. In the survey's judgment, surprising behavior does not establish a machine will, and persisting as a lineage, tending to serve one and remaining under unbroken control are different properties.

what ASK ruled

ASK decided this question only for the readings and axes stated:

• On the conditional-ability reading: yes, a constrained will can be free.
• On categorical ability, with the past and the laws held fixed: not settled.
• On the leeway (PAP) axis: bounded discretion is an analogue, with a recorded break.
• On the sourcehood and hierarchical-control axis: an analogue, with a distinct recorded break.
• As to instance: no. Bounded discretion is not an instance of metaphysical freedom.

ASK ruled these states and ruled no reason for them. It also ruled that the survey's pairing of bounded discretion with Ronald Dworkin's distinction remains unclassified, with no relation such as instance or analogue assigned to it. The ruling does not call bounded discretion incompatible with metaphysical freedom.

what remains open

ASK left categorical ability unsettled, and the survey records the consequence-argument and Frankfurt-case disputes as unresolved. The survey treats the separability of leeway and sourcehood as live, and finds no relation established among Frankfurt's identification, Ronald Dworkin's weak discretion and bounded discretion, each pair resting on a single unreplicated reading compared over summaries. It also records that the gloss of the strong sense in ASK's normative-apex doctrine, as the authority to set the standard, diverges from Ronald Dworkin's text.

Question 6 // Is the interface/surface the metaphysical boundary between minds, or a rendered relation among agency, delegation, authority, and perceived personhood?

what it asks

Other minds are ordinarily met through a surface: behavior, speech, a screen. The question sets two readings of that surface side by side. On the first, the interface is where one mind really ends and another begins, a boundary in the nature of things. On the second, what appears there is rendered, meaning produced by a set of relations: who is acting, who delegated to whom, who holds authority, and whether an observer perceives a person. The survey records that the question is not itself about who may hold the source-of-intent role, but bears on how that would be argued. It connects to ASK's method idea of a seed-mediated generated interface, in which an AI runtime renders a working interface from a durable seed.

the main positions

No individual source is recorded as evidence on this question, though material with candidate bearing on it exists, as described below. The survey pairs its theme with three candidate literatures: the other-minds problem (how anyone can know that another being has a mind), social cognition (how we come to attribute awareness to others), and perceived agency and the intentional stance (interpreting a system as though it had beliefs and desires). It records that pairing in the list where it sets ASK's ideas beside the literature only as a candidate, because the material behind it has not passed the survey's source-tracing rule.

The consciousness part of the survey does contain material in these literatures. That material was gathered, however, without recording which source supports which statement. Under the survey's rules it may be described and discussed, but no statement from it can carry weight in a recommendation until it has been traced to an exact source, and none has been. The survey's later work on free will, on the links between consciousness and free will, and on machines, personhood and recourse added nothing that could carry weight.

The survey also carries two guards. Explaining when people attribute consciousness shows neither that those attributions are correct nor that they are false. And because a system trained on human reports can reproduce the form of first-person reporting, its fluent reports no longer discriminate and establish neither consciousness nor standing.

ASK's own conclusion, set out separately on this page, holds that behavior at an interface cannot by itself establish the nature of the process producing it, or whether that process is conscious.

what ASK ruled

On August 23, 2026, ASK ruled nothing on the question itself. ASK adopted a three-part statement of the evidence.

1. Material on interfaces and surfaces exists in the consciousness part of the survey.
2. It has not passed the survey's rule that each statement must be traced to its source, and it carries no weight in any recommendation.
3. No admissible statement currently decides whether the interface or surface is a metaphysical boundary between minds.

Because nothing admissible decides it, the whole question stays open. ASK opened no further step on it.

what remains open

The whole question. The survey's own analysis proposed a statement-by-statement tracing pass over a defined set of entries as the next step, if one were opened. That is a recommendation and carries no authority. Tracing the entire consciousness material would mean re-reading its sources, which is new research that nothing authorizes.