NOTE-130: Floridi — The Informational Nature of Personal Identity

Contribution

The paper recasts the unity of the self as an engineering question about a multiagent system (“the problem of the chariot”) and argues that identity over time is not a proper metaphysical question at all, only an LoA-relative, purpose-driven one. It proposes the three-membranes (3C) model as a constructive answer to the synchronic unity question that Hume declared insoluble and Kant answered only transcendentally. It then uses the model to classify the effects of ICTs on the self (embodiment, presence, ageing, memory, the “digital gaze”) and places selves inside informational structural realism as the structures through which “information … becomes conscious”.

Key insight

For Floridi, “same person” questions have no answer until a purpose fixes the observables. Saul and Paul, the caterpillar and the butterfly, Theseus’s ship “are and are not” the same, depending on the level of abstraction (p. 553). What does have a substantive answer is what makes a self one thing at a time. That answer is layered binding: chemistry holds a body together, mutual information a cognitive system, shared semantics a mind. Anything that acts on those bindings, including information technology, is literally a technology of the self.

Assumptions

Premises and authorities.

• The method of levels of abstraction (Floridi 2008c): any system, the self included, “is always accessed and hence modelled at a given level of abstraction” (p. 553). This is assumed, not defended here, and it is what licenses the deflation of diachronic identity.
• Informational structural realism (Floridi 2008b = LIT-152; 2011b): the world is “the totality of informational structures dynamically interacting with each other” (p. 564). This is stated as a view defended elsewhere, and the paper’s closing claim is made conditional on it (“If this is the case—or at least in order for a philosophy of personal identity to be consistent with such a view”).
• Endurantism and perdurantism presuppose “direct realism”, a “non-mediated access to the system in itself” (p. 553). This characterization is asserted.
• No “substantialism, essentialism, or dualism is presupposed” about the self (p. 550).
• Supervenience: each membrane “builds on the previous one” (p. 558).
• Kant’s transcendental argument establishes that the self’s unity must be presupposed, which shifts the burden of proof (p. 556).
• Wittgenstein’s private-language argument: no private semantics, hence the consciousness membrane is a social achievement (p. 560). It is then partly reversed for embodiment: once a public language exists, it may be “privatised”, like the ladder (p. 561).
• Ockham’s razor is used to posit the same mechanism of encapsulation at all three stages (p. 558).

Key results

What the paper argues, by section.

• The chariot (pp. 551–552). Plato’s charioteer and two horses form an engineered multiagent system (MAS). This licenses a “constructionist or design-oriented approach”, on which MAS challenges (communication, coherence, coordination) “are just AI translations of classic issues in the philosophy of the self”. The hidden fourth component, the chariot and tack, stands for what “guarantees the unity and coordination of the system”. That is the problem of the chariot, and it is also an instance of Theseus’s ship.
• Two branches of egology (pp. 552–554). Diachronic egology concerns identification through time or across possible worlds. Synchronic egology concerns individualisation in time or in a world. The paper pursues only the latter. For the former it sketches an LoA approach: identity (“a is this”) and sameness (“this is the same a as that a”) depend on the observables fixed by the question’s purpose (the hospital that became a school: same by “location”, different by “social function”). “This is not relativism: given a particular goal, one LoA is better than another” (p. 554). Locke was right to distinguish kinds of sameness question and wrong to commit ontologically.
• Individuation first (pp. 554–556). “individualisation logically precedes identification” (p. 554). Lockean (consciousness, memory) and Narrative (Schechtman 1996) accounts both individuate through information processing, so both presuppose informational agents. Hume’s Appendix confession (the bundle has no principle of connexion) states the chariot problem for the informational self, and Kant’s transcendental unity is non-constructive. Floridi’s own earlier informational self (2005b, 2006, on privacy) left the same question open (p. 556).
• 3C model (pp. 557–560). Three phases of “auto-structuring” encapsulation, each placing “more not less distance” between the entity and its environment:
  • Corporeal: a “hardwired” divide. Its ontological function is inside and outside; its negentropic function is to resist the second law; its epistemological function is selective permeability. Data are transduceable physical patterns; stability is physical homeostasis. “A good example is a sunflower.”
  • Cognitive: “semi-hardwired (because configurable)”. Data become encodable resources with a direction from sender to receiver (“vectors” rather than “scalars”) and Shannon’s model applies; stability is cognitive homeostasis (memory and language). “a bird on the sunflower”.
  • Consciousness: “softwired (programmable)”. Data become “repurposable information, including conventional meanings”; the body becomes the outside of an inside experience; stability is mental homeostasis. “a gardener watching the bird on the sunflower”. The self is indivisible because self-division yields “mere schizophrenia”, not two selves (p. 560).
  • Bonding forces (p. 560): chemical bonds, then mutual information, then semantics. “The problem of the chariot therefore may be solved only by taking into account all the bonding forces”.
• Virtual machines (n. 6, p. 558). The encapsulations involve “increasing … virtualisation”, which is “nothing metaphorical”, but Floridi also dismisses “virtual functionalism” (Sloman; Pollock 2007) as no more instructive than classic functionalism: “virtual machines generate virtual problems about virtual minds virtually conscious”.
• ICTs as technologies of the self (pp. 560–563). ICTs “re-ontologise” self-poiesis by enhancing the corporeal, empowering the cognitive and extending the consciousness membrane. Once formed, the conscious self is “conceivably independent” of the specific body that produced it (the springboard). It needs some platform but not just any, which makes “eterobodiment” and mind uploading conceivable (p. 561). The self is “located in the brain but not present in the brain” (p. 562). Digital selves “outdate” rather than age. Accumulated records freeze the self, so “Forgetting is also a self-poietic art” (p. 562). The “digital gaze” re-ontologises selves, sometimes “eteronomously” (p. 563).
• Logic of realisation (pp. 563–565). Anagnorisis is the process by which “a later stage in the information flow … forces the correct reinterpretation of the whole information flow”. It is not falsification, because earlier information stays true but is re-framed (fraternal love). Applied to the 3C story, it is the self narrating its own emergence. Under ISR, “Selves are the ultimate negentropic technologies, through which information temporarily overcomes its own entropy, becomes conscious” (p. 564). “There are still only informational structures”, some of them minds, and “Only minds are able to interpret other informational structures as things or organisms or selves” (p. 565).
• Conclusion (p. 565). A call for an “ethics of self-poiesis”, an “ecology of the self”.

Claims

id	claim	strength	support
C1	Diachronic identity and sameness of selves are relative to a purpose-fixed LoA, and LoA-free metaphysical versions make no sense	informal argument from examples	pp. 553–554: hospital/school, Saul/Paul, Theseus’s ship. Rests on the LoA method (Floridi 2008c), assumed
C2	This is not relativism, because a goal ranks LoAs	assertion	p. 554, one sentence
C3	Individualisation logically precedes identification	assertion	p. 554: an appeal to Plato in three sentences
C4	Lockean and Narrative accounts both individuate informationally and inherit Hume’s unity problem	informal argument, textual	pp. 555–556, long quotation from Hume’s Appendix
C5	Kant’s transcendental argument solves the unity problem only partially, because it is non-constructive	informal argument	p. 556
C6	The self is individuated by three successive encapsulations (corporeal, cognitive, conscious), each bonded by a distinct force (chemical, mutual information, semantic)	hypothesis, self-described as “somewhat figurative”	pp. 557–560. No criterion is given for when a bond or membrane is present
C7	The same encapsulation mechanism operates at all three stages	assertion via Ockham’s razor	p. 558 (“we should take seriously the possibility”)
C8	Selfhood requires a community, since semantics cannot be private	informal argument	p. 560: Wittgenstein
C9	Once formed, the conscious self can in principle run on a different platform (though not any platform)	informal argument by analogy	p. 561: springboard, butterfly, Wittgenstein’s ladder
C10	ICTs are technologies of the self because they affect each membrane	follows from C6 by definition	p. 560: “any technology capable of affecting any of them is ipso facto a technology of the self”
C11	Recorded memories reduce the self’s freedom of self-definition	assertion with illustration	p. 562
C12	Under ISR, selves are the stage at which information becomes self-aware	conditional on ISR, assertion	p. 564

Concepts

• egology: the theory of the self; diachronic egology is identification through time or worlds; synchronic egology is individualisation at a time (pp. 552–553).
• problem of the chariot: what makes a multiagent self a coherent unity and source of action, and keeps it so (p. 552).
• level of abstraction (LoA): a set of observables fixed for a purpose; “an epistemic interface” (n. 4).
• identity / sameness: “a is this” versus “this is the same a as that a”, both LoA-relative (p. 553).
• membrane: a self-assembled, self-organised divide between a new inside and outside; corporeal (hardwired), cognitive (semi-hardwired), consciousness (softwired).
• egopoiesis / self-poiesis: the construction of the self; ICTs are egopoietic technologies (pp. 550, 560).
• re-ontologise: change the very nature of something (pp. 561, 563).
• eterobodiment: the self as a cross-platform, not platform-free, structure (p. 561).
• anagnorisis / realisation: information that forces reinterpretation of an entire information flow without falsifying it (p. 564).
• I before Mine (IBM): the self appropriates and unifies bodily and cognitive events as its own (n. 7, from Floridi 2005a).

Connections

Metaphysical thesis defended or presupposed. It defends two metaphysical theses and presupposes a third. (1) An anti-metaphysical thesis about diachronic identity: persistence questions about selves are LoA-relative and purpose-bound, with no LoA-free fact. (2) A constitutive thesis about synchronic unity: a self is an emergent (supervenient) informational structure individuated by nested encapsulations with increasingly virtual bonding forces. (3) The presupposed backdrop is informational structural realism (LIT-152), on which “There are still only informational structures”. The metaphysics tag is justified. identity is the right primary topic, with the caveat that the paper’s notion of identity is individuation and unity, not persistence. The consciousness tag is defensible (the third membrane, self-awareness), though the paper offers no theory of consciousness.

Links in this record.

• LIT-152 (Floridi, A defence of informational structural realism) is the ontology this paper assumes, cited as 2008b; LIT-215 extends it.
• LIT-130 (SEP Personal Identity and Ethics): the mainstream persistence debate (psychological vs biological criteria, Parfit’s “what matters”) that Floridi declines to enter. His claim that identity questions “are really teleological questions, asked in order to attribute responsibility … make sense of one’s own life” parallels the SEP entry’s practical-concerns approach, but he draws a deflationary conclusion where the entry treats the metaphysics as open.
• LIT-097 (The Extended Mind): Floridi’s “eterobodiment” and memory-externalisation claims sit near extended cognition, but he locates the self inside a “consciousness membrane” that detaches from the environment rather than extending into it.
• LIT-206 (Goldstein & Lederman, instance agent): a precise answer to an individuation question for LLMs, where Floridi’s LoA method would say the answer depends on the purpose.
• LIT-138 (The Depth of the Body): both treat an information-level object as a genuine constituent of a person.
• LIT-005 and LIT-049 (Bourrat on information-theoretic and coarse-graining individuality): formal versions of the idea that mutual information marks off an individual from its environment, which is what Floridi’s cognitive membrane asserts informally.
• LIT-154 (Leitgeb & Ladyman, criteria of identity and structuralist ontology): structural realism’s own individuation problem, which ISR inherits and this paper does not address.

Bearing on the record

It carries nothing for ML practice, and there is no Anthology document it should inform. Its MAS reading of the self and its “AI translations” remark are programmatic, not technical. For nucleation it is a primary source for the level-of-abstraction deflation of identity questions, which is directly relevant to questions about copying or forking AI instances. The dossier’s framing, a foil to the extended mind and to instance-agent individuation, holds, but the record should cite it as a view about individuation and unity with a deflationary stance on persistence, not as a criterion of personal identity.

Limitations

• The 3C model is described as “somewhat figurative” by its author (p. 557). It gives no criterion for when a membrane or bonding force is present, so it cannot adjudicate a disputed case, such as whether an AI system has a consciousness membrane.
• The individuation-first thesis (C3) is asserted, and the deflation of diachronic identity (C1) rests on the LoA method, which is assumed from elsewhere.
• Characterising endurantism and perdurantism as presupposing “direct realism” is asserted without engagement with either literature.
• The mapping of bonding forces to membranes (mutual information for cognition, semantics for consciousness) is not argued, and “semantics” as a bonding force is not defined.
• Having set diachronic identity aside, the paper still draws diachronic conclusions: the self can survive on another platform (C9), and memories fix it over time (C11).
• The ISR conclusion (C12) is conditional on a view defended elsewhere, and its evidential support here is nil.
• The ICT illustrations (2011 examples: Facebook, Diaspora) are dated, and several are observations rather than consequences of the model.

Open questions

• Can the 3C model be given membership conditions (e.g. via information-theoretic individuality measures of the kind in LIT-005 and LIT-049) so that it classifies cases?
• Does the LoA deflation of diachronic identity survive cases where purposes conflict, such as responsibility versus prudential concern? The paper says one LoA is better “given a particular goal” but gives no way to adjudicate between goals.
• What would the ethics of self-poiesis called for in the Conclusion require? The paper leaves it as a programme.

Corrections to the seeded skim

• The dossier, LIT summary and abstract present this as an account of personal identity. The paper explicitly sets the diachronic (persistence, re-identification) question aside and deflates it. Sameness relations “are satisfied according to the LoAs adopted”, diachronic identity questions “are really teleological questions”, and “insofar as they are dealt with metaphysically (modally or not, it does not matter), they do not deserve to be taken seriously” (pp. 553–554). The paper is about synchronic individuation (unity), not persistence. This is the reading’s most important finding and neither the dossier nor the LIT states it.
• The dossier says Lockean and Narrative accounts are “treated as needing a prior answer to what the self is”. The text treats them as individuation approaches that both presuppose informational agents (p. 555), and whose shared failure is Hume’s: they cannot explain what binds the information into a unity (pp. 555–556). Kant’s transcendental unity is a “partial, epistemological” and “non-constructive” answer (p. 556). The 3C model is offered as the constructive answer that Hume and Kant lacked.
• “Individualisation logically precedes identification” is not argued beyond three sentences appealing to Plato (you cannot look for something without knowing what you look for, p. 554). The dossier rightly calls it load-bearing; it should also say it is essentially asserted.
• The dossier omits the bonding forces, which are the model’s actual answer to the chariot problem (p. 560): chemical bonds (corporeal), mutual information (cognitive), semantics (consciousness). It also omits the consequence Floridi draws: because semantics cannot be private (Wittgenstein), a self needs a community, and “the self emerges as a break with nature, not as a super connection with it” (p. 560).
• The dossier says the model “detaches an inside from an outside” at each stage. That is right, but Floridi’s hedging is missing: the model is “biologically and informationally plausible, but … admittedly, somewhat figurative”, and its aim is “ultimately” to explain why ICTs are egopoietic (“I hope that the model at least achieves this much”, p. 557). The unity is “achieved, or at least described” by it.
• Minor: the section for the first membrane is headed “Phase One: The Physical Membrane” though the text calls it corporeal; the abstract calls the third component “the conscious”. The Plato quotation is cited as “Phaedrus 246a–254e”.
• metaphysics tag: justified — the paper defends a thesis about what selves are (informational structures, emerging by supervenient encapsulation) within informational structural realism, and it takes a metaphysical position against LoA-free identity questions.