The day can change her.
A score cannot define her.

Back at home, Mira looks at the presentation again. The morning was demanding, but she would choose it again. A system optimised only to lower arousal might have suggested avoiding the very event she values. Her reflection brings in a different kind of context: what she is trying to become, what discomfort she considers worthwhile and which judgements she wants to retain the authority to revise. These questions concern an evolving person, rather than the interpretation of a single signal.

This is the role of ~WellbeingIdentity in the IoV architecture. Its nodes include given and earned identity, platform-based rented identity, skills, a moral compass, story, living identity state and consent. The vocabulary extends beyond physiology into biography and agency. It remains a representation of a person, with the limits and governance obligations that any representation introduces.[13]

Version 0.7.0 also distinguishes a calculated daily Wellbeing state from a constituted Genesis Baseline. Baseline formation draws on identity, observed Wellbecoming, first-person assessment, consent and issuer review; it is not simply a sensor average. Its replay boundary prevents pre-Genesis observations used as baseline evidence from later being counted again as movement. The distinction is a specification contract, rather than evidence of clinical validity.[14]

The difficult question turns back on my own work. IoV includes a composite ~~~WellbeingScore under ~~IdentityState. What construct does it represent, how sensitive is it to missing observations, and when could a mathematically consistent change misrepresent someone’s life? A name in a registry cannot answer these questions any more than the name Entropy can validate a product score. The obligation to define assumptions and test usefulness travels with both systems.[15]

Control over the record is equally consequential. Temple’s vision states that conversations are encrypted and do not reach a model company. Its privacy policy, updated 31 October, permits certain app features to use third-party AI with prompts and device metrics, while prohibiting those partners from using the data to train or improve their services. These may describe different processing paths. The documents leave a practical question about which path handles which input; they do not justify announcing a proven contradiction.[9][16]

For Mira, useful answers would identify where her spoken note is processed, what is retained, what she can export and how a correction propagates to later inferences. Temple’s policy promises access, correction, deletion, export and consent withdrawal. Those promises can be examined in implementation. IoV’s consent language needs the same practical examination: a disclosure rule matters only if the systems carrying the record honour it.[16]

At dinner, the menu never claimed to own Mira’s review or decide whether she was a discerning person. Technologies that build persistent profiles can acquire far more power. Representing her well therefore requires both richer evidence and boundaries around its use. A better account of a human being should leave that human able to challenge the account, including when the record looks orderly and the model sounds certain.