Administrative identity is not what matters

What would philosopher Derek Parfit make of the Integrated Data Infrastructure?

Feb 24, 2026

Editor’s note: My sincere apologies for the break in service over the summer months. Work and other commitments got the better of me. This post marks a resumption of normal service — Dave.

In a terrible accident, Derek suffers fatal injuries to his body while his brain is unharmed. His two brothers are also involved in the accident, but suffer different consequences; they are brain-dead with healthy bodies. Faced with the aftermath of the accident, surgeons split Derek’s healthy brain in half, and put each half in the body of his siblings. When they wake up, the two resulting people believe they are Derek, remember living his life, and have his personality. Their bodies are similar to Derek’s before the accident.

Blue duck *Hymenolaimus malacorhynchos*, Ruahine Forest Park *Chris Tuffley*

One identity, two instances

What became of Derek? Did he die, or did he double? Is one brother him and not the other? For whom does his family hold a funeral? This thought experiment is one of several by philosopher Derek Parfit, compiled in his epic tome Reasons and Persons (1984).

Parfit concludes there would be no one identical to him after the surgery. “There will be two future people, each of whom will have the body of one of my brothers, and will be fully psychologically continuous with me, because he has half of my brain,” he writes. If identity mattered, he argues, then this result would be just as bad as death. But the result cannot be as bad as death, since he has two psychologically continuous selves. “Since I cannot be one and the same person as the two resulting people, but my relation to each of these people contains what fundamentally matters in ordinary survival, the case shows that identity is not what matters.”

Over the course of several chapters, Parfit methodically explores the ramifications of this conclusion. He argues that what matters is psychological continuity and connectedness, which is a relation of degree; you might feel strong psychological connectedness between yourself today and tomorrow, but weak or non-existent connection with yourself in ten years’ time.1 Given that, over sufficient time, psychological connection with ourselves can cease, connectedness can be stronger with other people in the present than with yourself in the distant future.

Parfit acknowledges there is a difference between logic-ing your way to a conclusion and actually feeling within yourself that it holds truth, but he encourages us to stick with it:

When I believed [that personal identity is what matters], I seemed imprisoned in myself. My life seemed like a glass tunnel, through which I was moving faster every year, and at the end of which there was darkness. When I changed my view, the walls of my glass tunnel disappeared. I now live in the open air. There is still a difference between my life and the lives of other people. But the difference is less. Other people are closer. I am less concerned about the rest of my own life, and more concerned about the lives of others.

One identity, one unique identifier

On the third floor of Victoria University of Wellington’s Rutherford House, there’s a swipe-card-protected door which only opens to those who have been granted approval for their research by Stats NZ, hold good-enough coding skills, have undergone online training in confidentiality, and have signed a legally binding lifetime commitment to keep the data they encounter confidential.

The privileged room in Rutherford House is one of many datalabs across New Zealand, which quietly capture space in government offices, universities and research institutions. Each datalab has its own quirks, some with desks so close together that researchers knot themselves up to slot in the empty chair, while others have carefully placed blinkers between desks as a constant reminder of the sanctity of the space.

When aerial window cleaners set up ropes around Rutherford House and began bungeeing past the third floor, datalab use was only permitted with the blackout blinds fully extended, to prevent curious eyes outside from spotting a morsel of confidential information. On that specific occasion, the risk was very, very small. The chance of someone being able to see and interpret information on a computer screen several metres away through glass glistening with soap suds, while researchers, sitting inches from the screen, strained to make sense of spreadsheet cells, red error messages, data dictionaries and metadata, was small. But it was not zero, and it could be reduced to zero, and so it was.

The Integrated Data Infrastructure (IDI) is a web of data on the people of New Zealand, which can only be accessed in these locked rooms. I first used the IDI in 2017 as a PhD student and have since used it in fits and starts, as and when my workplaces needed it. Administrative data, the residue left behind from every interaction with the state, is collected by government departments and sent to Stats NZ, which strip the names and replace them with that person’s unique identifier, a stream of numbers about eight to ten digits long. Correctly managed, a person’s number is assigned to them each time their data arrives at Stats NZ, so the information from each source can be linked at the person level.

When I think about myself-as-administrative-data, I can easily imagine my information; my university course enrolment and grades, each time I left and entered the country, every medicine I picked up from the pharmacy, my visas, monthly earnings, ACC reimbursements, my census responses, and my state-recognised partnership, which links me to my anonymised partner and all his data too.

I get excited about the potential this data holds, for really saying something; to reveal some kind of truth which holds for a group of people at a specific point in time. But when it comes to actually carrying out research, administrative data drags significant challenges along with it. I have to constantly remind myself that this data is subjective, an abstraction. The variables recorded are not conceived of by a divine, omnipotent statistical overlord, but iteratively, sometimes accidentally, for operational purposes.

The one part I never thought to question was the unique personal identifier.2 A person is a person is a person; one of the few essential truths the IDI can stake a solid claim to. That is, until I read Derek Parfit’s Reasons and Persons. If I am, psychologically speaking, closer to other people now than I am to myself in the future, why are individual people set in individual grooves for our entire administrative lives, as if our continuities are defined by national health identifiers (NHIs), passport or IRD numbers? What would happen to Derek’s unique identifier in the event of the accident and surgery? Is there a person who decides?

I’m not arguing for the removal of the person-level unique identifier; the IDI is not for making metaphysical claims about the nature of self, but for research, and for research purposes it’s helpful to simplify a person down to their bodily boundaries. I actually think Parfit would have loved the IDI. He was fundamentally concerned with representation, always asking if he was looking at the thing itself or a copy, and to what extent they are the same.

The IDI is a fantastically big representation of ourselves and it does matter, in that it offers opportunity for knowledge. But it is not what matters. Just as the IDI can’t fulsomely capture what matters from a research perspective — happiness, safety, social connection, purpose — neither can it capture what mattered to Derek Parfit; overlapping chains of memories, beliefs, intentions and characteristics, and the strength of those links.

While waiting for my inelegant code to process, I imagine what Parfit would do if he got a research project approved, passed confidentiality training and learned SQL,3 and was on the desk next to me in the datalab. Would he combine the unique identifier and time variables to create a scale for each person, indicating how different any record in the past is from their present? Would he have an alternative to the individual fixed effect I’ve been using all these years under the assumption that a person is consistent over time? What research questions would he answer with his alternative coding of the individual, and would his results improve the wellbeing of New Zealanders?4

One identifier, zero instances

Derek Parfit died in 2017, in London, aged 74. His death would have been confirmed by a health professional, and entered into a spreadsheet; his UK National Health Service identifying number would have been deactivated and never issued again; the death of an administrative identity. For Parfit, his death was more a question of degrees.

When I believed [that personal identity is what matters], I also cared more about my inevitable death. After my death, there will be no one living who will be me. I can now redescribe this fact. Though there will later be many experiences, none of these experiences will be connected to my present experiences by chains of such direct connections as those involved in experience-memory, or in the carrying out of an earlier intention. Some of these future experiences may be related to my present experiences in less direct ways. There will later be some memories about my life. And there may later be thoughts that are influenced by mine, or things done as the result of my advice. My death will break the more direct relations between my present experiences and future experiences, but it will not break various other relations. This is all there is to the fact that there will be no one living who will be me. Now that I have seen this, my death seems to me less bad.

Administrative systems require fixed identity, but Parfit reminds us that what morally matters may not map onto those fixed units.

By Olivia Wills

I read some old emails yesterday, which date back to 2013. They were sent from my email address, but they certainly weren’t written by me; there is nothing in them that resonates with my current self, no style, no content, and I have no memory of them. As you can imagine, they are mortifying and I gladly reject any notion they might have been written by “me”.

I mean this in the metaphysical sense. A unique identifier wrongly assigned makes me, like all good data analysts, ask many questions.

SQL, typically pronounced “sequel”, is an acronym for Structured Query Language. It is a computer programming language for extracting data from large datasets, and for manipulating such data.

Elsewhere in Reasons and Persons, Parfit argues against the use of a discount rate. Fortunately for all of us, that is not the topic of this essay. On that topic, see: