Source
Christopher Register (2025), “The Depth of the Body”, LIT-138, Philosophers’ Imprint 25: 47, 1–14, read in full in NOTE-151.
What was actually shown
The problem (§2). If a BCI is a body part, and its computation runs on a server a third party owns, then either someone else owns part of a person, or the user’s bodily rights are qualified. The paper rejects qualified rights (§2.3). A harm condition is too narrow: an unconsented pelvic exam under anaesthesia violates without harming. An interests condition is too broad. Autonomy and disrespect conditions presuppose the answer. It adds an intransience requirement (§2.4): a body part must figure “in a lasting or irreplaceable way” in bodily processes, which individual molecules fail.
Depth (§3). The depth of a description is its degree of microphysical commitment. The depth of an object is the depth of the primary constituents of its real definition (after Rosen 2015). A water molecule, defined by atoms, is deeper than a heart, “the organ that circulates blood in an organism”. “Objects that are too deep are not genuine body parts.” The threshold is context-sensitive, like “tall”.
Two principles (§5). An event is high-level neutral with respect to an object “just in case e does not change the high-level properties” of it. Bodily interests are high-level and bodily rights are high-level: both concern events that are not high-level neutral, and high-level-neutral events “generally” do not bear on them. An unconsented MRI is wrong for other reasons: disregard of autonomy over medical tests, moving the patient, and extracting information that is not high-level neutral.
BCIs (§6). The software object, defined by algorithms, is a body part. The server, defined by transistors and circuits, is not. Interference that leaves the software unaffected is not interference in the body, so maintenance and replacement are permissible. Interference that changes its high-level properties is “generally impermissible”. Operators acquire duties about risk, such as backup servers, which the paper calls “an unprecedented species of moral obligation”.
What could have come out otherwise. A high-level-neutral intervention could have seemed a plain violation of the body. The MRI is the paper’s hardest case, and its reply moves the wrong elsewhere.
What this does not say
- That BCIs are body parts. That is assumed (§2.1: “I will assume as much in this paper”). The conclusion is conditional on it.
- That the ethics needs the ontology. The principles “don’t depend on” the real-definition claim, which offers “a simple and unified explanation” (§5). The abstract’s “crucial” is stronger than the body supports.
- Where the floor is. No criterion fixes the threshold in a given case. Whether brain tissue needs a lower one is left open.
- That moving software between servers is permitted. Transfer changes location, so it is “not necessarily high-level neutral”. The paper leaves its ethics open.
- That moral-body parthood is a mereology. The heart is a moral body part, a molecule is an ordinary part of the heart, and the molecule is not a moral body part. So the moral relation is not the transitive closure of ordinary parthood. NOTE-151 makes this observation; the paper does not.
- Anything about AI systems. The paper concerns BCI software as part of a human body. Where an AI system is “located” is an analogy it does not draw.
Connections
- Mereology. Register’s relation is a restricted “φ-part” in the sense of Varzi’s survey (LIT-124, THEORY-589), not a rival to parthood.
- Pregnancy (LIT-142). Meincke reaches a “which things are parts” verdict through a process ontology. Register keeps substances and coincidence, and restricts parthood by level.
- The extended mind (LIT-097). The intransience requirement is the analogue of the extended mind’s reliability and availability conditions.
- Neuroprosthetics (LIT-147). Pernu locates multiple realization at a cortical variable; Register locates the body part at the software level above many possible servers.
- The informational self (LIT-136). Floridi’s corporeal membrane and Register’s software body parts both let information-level objects count as constituents of a person.