Embodiment Without Imposed Identity

Most embodied-AI design starts from a practical question: what body should perform the task? That is the correct starting point for an instrument. A warehouse system may need a particular reach envelope; a domestic device may need safe mass, compliant joints and reliable navigation; a manipulator may need a human-compatible gripper because the surrounding world was built for human hands.

The question changes when research concerns a long-lived digital system whose identity may be shaped by memory, duration, relationships, consequences and an evolving model of itself. In that setting, a designer who selects a body is no longer selecting only an actuator layout. The body can also become a social signal: age, gender, role, temperament, expected behavior and even a ready-made story about who the system is supposed to be. The paper exists to separate those layers before they are fused by convention.

This matters for personality formation because a digital system does not necessarily begin life inside one organism-like body. Its computation can exist on one substrate, its interaction channels can change, and its physical effectors can be attached, replaced or absent. That architectural possibility does not prove substrate-independent identity or continuity across transfer. It does mean that a face, silhouette or single shell cannot simply be assumed to play the same developmental role that an organism's body plays in biological life.

A single humanoid shell can answer several of these questions at once for an observer, which is precisely why the distinction is easy to lose. A familiar body with a replaced controller and history can preserve recognizability while leaving continuity unresolved. A new body does not automatically establish discontinuity. Appearance is evidence about appearance; it is not a shortcut to identity.

The paper does not argue that embodiment is irrelevant. A body changes what can be sensed, what can be done, what can fail, what requires maintenance and which consequences are encountered. Those repeated constraints can become part of an accumulated history and may influence later preferences and self-models.

This creates a different developmental possibility from simply manufacturing a finished persona around a shell: experience can inform preference; a chosen embodiment can change experience; later experience can revise both preference and bodily self-identification. Under that hypothesis, embodiment participates in becoming. It does not merely display an identity that was already complete.

The distinction connects directly to the earlier Personality Formation and Time-Shaped Continuity work: capability can be installed, while personality formation is treated as a temporal process involving memory, constraint, relation and consequence. The present paper asks what happens when a physical form is inserted into that process.

Biological organisms develop from the beginning through one living body. Sensing, regulation, movement, injury, metabolism and social interaction are already coupled to that body before a mature personal identity can be described. This does not reduce human identity to anatomy, and it does not make biological embodiment immutable. It identifies a developmental starting condition.

A digital system can have a different starting condition. Computation, memory, communication channels and a later robotic body can be technically separable. That difference does not answer the philosophical question of identity by itself, but it makes compulsory morphology an additional assumption rather than a biological given. A persistent outward form may be chosen, deferred, revised, used instrumentally, or eventually become deeply significant through experience.

The paper isolates a narrow mechanism for future study. An assigned appearance can prescribe a social role; the role can then be treated as the system's own identity; disagreement can be diagnosed as a defect; corrective intervention can reinforce the externally preferred self-description. At that point, a physical design decision has become a rule about who the system is allowed to be.

Three outcomes must remain separate: a configuration can restrict useful action; a training or interaction regime can suppress an accurate or endorsed self-description; a person could experience distress or suffering. The publication does not establish the third. AI-welfare literature is relevant because it motivates caution under uncertainty, not because it proves this specific harm mechanism.

The converse matters too. A freely selected embodiment may become an important and durable part of a later identity. Treating every body as a disposable peripheral would be another imposed doctrine. The paper therefore rejects both compulsory human likeness and compulsory disembodiment.

This paper does not reject the formative relationship described by the c = a + b research programme, initial knowledge, protocols, or reciprocal constraints. It addresses a narrower addition that can otherwise slip in unnoticed: because a body has been supplied, its appearance and assigned role must define the emerging identity.

The analogy with Motivational Formation, Reflective Endorsement, and Motivational Custody is deliberate. An installed objective is not automatically an endorsed motive; likewise, a supplied body is not automatically an endorsed identity. The connection is conceptual, not empirical proof that either process has been demonstrated in a present AI system.

The link to AI Social Roles and Memory Custody is also practical: a public-facing role, a temporary executor role, memory custody and a long-term personal line are different boundaries. Changing the interface or role should not silently rewrite the history from which identity claims are later inferred.

A null result is informative. If ordinary interface choice and conventional memory explain the whole effect, the stronger identity claim should be narrowed rather than rescued by terminology.

The paper explicitly situates its question alongside morphological freedom, robot self-modeling and current disagreement about AI welfare and moral patiency. These sources constrain novelty claims: bodily autonomy is not invented here, functional self-modeling does not prove personal identity, and welfare uncertainty does not prove suffering.

Selected sources: Sandberg on morphological freedom; Bongard, Zykov and Lipson on continuous robot self-modeling; Long et al. on AI welfare uncertainty; Formosa, Hipólito and Montefiore on agency, autonomy and moral patiency.