Things change.
Members leave. Components are replaced. Records move. Rules are amended. Leaders disappear. Software is rewritten. Communities lose their platforms. Institutions outlive their founders. Cultures pass through generations. Artificial systems change models while retaining memories, instructions, relationships and histories accumulated elsewhere.
Eventually the same question appears:
Is this still the same thing?
It sounds simple until an answer has to be justified.
The usual evidence is weaker than it first appears. The name remained. The files survived. The same people are present. The legal entity still exists. The database was migrated. The interface looks familiar.
Any of those things may matter. None is decisive.
A company can preserve its legal identity while abandoning almost everything that once made it recognisable. A community can lose its platform, archive and name while its members experience little doubt that the community itself survived. A tradition can preserve every ritual while losing the commitments that once gave those rituals meaning. A software system can be rebuilt almost completely and still be treated as the same product.
So continuity cannot simply mean that enough old material remains.
The deeper problem is deciding what kind of survival matters.
Philosophy has worried at that problem for centuries. The Ship of Theseus remains useful because it separates material persistence from identity. If each plank is gradually replaced, does the ship remain the same ship? If the discarded planks are later reconstructed into another vessel, which one has the stronger claim?
The point is not that there is one accepted answer. There is not. Different theories privilege different relations: material continuity, causal history, function, biological persistence, psychological continuity, or some combination.
What matters is what the disagreement reveals.
Every claim of continuity already contains an assumption about what kind of thing is supposed to be continuing.
A ship, a person, an institution, a friendship, a community and an artificial agent do not necessarily persist in the same way. Their identity conditions need not be interchangeable.
That gives us the first important proposition:
Every continuity claim contains a theory of identity.
When we say that two states belong to the same thing, we are already deciding which changes matter and which do not.
Usually we make that decision silently.
The problem becomes more obvious when material persistence breaks down.
Consider an online community.
While the community remains inside one stable platform, it is easy to treat the community and the platform as though they were one object. Membership lists, administrators, messages, roles, permissions, reputation, history and conversation all occupy the same technical surface.
Then the container fails.
The group is deleted. Access changes. Moderation collapses. The service becomes unusable. The members deliberately leave. Whatever the cause, a replacement appears somewhere else.
New group. New server. New link. Perhaps even a new name.
People begin arriving, and something peculiar happens.
They recognise one another.
Old reputations still matter. Previous disputes remain intelligible. Some people are trusted immediately. Others are treated cautiously for reasons nobody needs explained. Shared jokes survive without footnotes. Informal roles begin reforming around familiar people. Norms that were never written down begin operating again.
At some point, somebody effectively says:
We’re back.
But what exactly came back?
Not the URL.
Not the server identifier.
Not necessarily the archive.
Not necessarily the whole membership.
Sometimes almost none of the original technical state survives.
Yet the participants behave as though something real crossed the boundary.
Research on platform migration and online communities supports that intuition, while also showing its limits. Communities can persist across technical environments, but migration can fragment relationships, alter network structures and change the community itself. The platform matters enormously because it shapes visibility, access, moderation, communication and memory.
But the platform is still not identical to the social system it hosts.
That distinction matters.
The room can disappear while the community survives.
The room can also survive while the community dies.
The question, then, is not simply what was preserved. It is what the preserved elements were doing.
Suppose every message from the old community is exported. Every profile survives. Every administrator badge is restored. Every channel is recreated. The archive is perfect.
But trust has disappeared. The old purpose no longer binds the group. Authority has become meaningless. Participants no longer understand the references they inherited. The relationships that made those records consequential have dissolved.
The information survived.
The community may not have.
Now reverse the case.
Most of the archive is lost. The old platform is gone. Only part of the membership migrates. The new group looks technically different.
But relationships survive. Reputation survives. Shared history remains intelligible. Norms still make sense. Participants still know what the group is for.
Very little data crosses the boundary.
Yet the community may continue almost without interruption.
That suggests a second proposition:
Continuity is not proportional to the amount preserved.
The quantity tells us surprisingly little.
What matters is whether the things that survived were the things that actually carried identity.
I use identity-bearing structure to describe those properties, relationships, commitments, histories, authorities or practices that matter to continuity for a particular kind of entity.
The phrase is intentionally broad because the contents will differ.
For a community, identity-bearing structure may consist largely in relationships, mutual recognition, reputation, shared history, norms and common purpose.
For an institution, mandate, legitimate authority, obligations, governance and succession may matter more.
For a culture, continuity may depend upon language, stories, practices and transmission.
For an archive, provenance and context are central.
For an artificial identity, candidates might include governing identity constraints, authorised memory, characteristic judgement, relationship history and the rules by which state is allowed to change.
There is no reason to assume these all share the same criteria.
But they share the same problem:
which differences can occur while this remains this?
Research on organizational identity helps here. Organizations can change substantially while their members continue to understand them as continuous. Work by Dennis Gioia, Majken Schultz, Kevin Corley and others has challenged the idea that organizational identity must be rigidly enduring. Identity can be interpreted, negotiated and adapted while continuity remains meaningful.
Historical institutionalism reaches a similar conclusion from another direction. Institutions can change gradually through layering, reinterpretation and conversion rather than clean destruction and replacement.
Social psychology adds another piece. Research on perceived collective continuity suggests that groups understand themselves through both cultural and historical connection. Present members experience themselves as belonging to something that extends beyond the current moment because they perceive links between current practices, inherited culture and earlier stages of the group.
That helps explain recognition.
But it does not settle continuity.
Because recognition can be wrong.
A familiar logo can manufacture reassurance. A copied interface can create false familiarity. Leaders can insist that nothing important has changed. Institutions can selectively construct continuity narratives. People can accept a successor because they want something to have survived.
Recognition is evidence.
It is not final authority.
The stronger question is:
Did the surviving elements preserve the grounds on which recognition was warranted?
That move matters.
Instead of asking only whether participants experience continuity, it asks what makes that experience defensible.
And that leads directly into another problem.
Suppose we reproduce every identity-bearing characteristic perfectly.
Same structure. Same language. Same practices. Same memories. Same stated purpose. Same behavioural patterns.
Does that make the result the original?
Not necessarily.
A replica may resemble its predecessor almost perfectly while having no historical relation to it.
So identity-bearing structure is not enough.
Continuity also requires lineage.
A later state must somehow arise from an earlier one through an intelligible relation of succession.
That relation may be causal, historical, institutional, social or governed, depending on the system.
Provenance offers a useful analogy. Archival practice has long treated records differently depending on their origin and context. The W3C PROV model similarly represents entities, agents, activities and derivations so that a later state can be connected to the processes that produced it.
Neither archival theory nor provenance modelling solves metaphysical identity. That is not their purpose.
But both make one thing clear:
history matters.
A thing is not merely its present configuration. How it arrived matters too.
Historical institutionalism makes the same point through path dependence. Earlier decisions shape later possibilities. Sequence matters. Timing matters. History is not decoration attached after the fact. It participates in what the system becomes.
That gives us a cleaner distinction.
Resemblance without lineage is replication.
And:
Lineage without preserved identity-bearing structure may amount to descent rather than continuity.
Continuity requires some defensible relation between the two.
This is the core of the framework:
Continuity requires both identity-bearing structure and lineage.
But lineage creates its own problem.
Something can possess excellent historical continuity and still inherit structures that should not continue.
An institution can preserve harmful authority arrangements. A family can reproduce destructive patterns in the name of tradition. A community can recreate the same hierarchy every time it migrates. A bureaucracy can continue a procedure for a century after everyone has forgotten why it existed.
Historical continuity does not confer moral legitimacy.
So inheritance and authority have to be separated.
A community can inherit a norm and reject it.
An institution can inherit a rule and repeal it.
A person can inherit a tradition and refuse it.
An artificial system could inherit previous judgments without being required to reproduce them.
Hence:
Inheritance is not obligation.
That is not an empirical finding. It is a normative proposition.
But it prevents continuity from collapsing into captivity.
If continuity requires preserving everything, then nothing is allowed to become different.
If continuity requires preserving nothing, then nothing remains accountable to its history.
A viable successor needs enough inheritance to remain intelligibly connected to what came before, but not so much that the past becomes permanent authority.
The past must remain capable of reaching the future.
The future must remain capable of refusing parts of it.
This is where the word legitimate becomes unavoidable.
Continuity cannot mean that nothing changed.
Living systems change precisely because they remain alive.
Organizations adapt. Institutions amend rules. Communities renegotiate norms. Cultures reinterpret practices. Software is updated. Artificial systems are modified.
Some changes may be necessary for continuity rather than threats to it.
The relevant question is therefore not:
Did this change?
It is:
How did this change, and why does that change still belong to the history of this thing?
Legitimate transformation does not necessarily mean morally good transformation. It means something narrower: the change occurred through processes compatible with whatever governs valid change for that kind of entity.
For a constitutional institution, that might involve valid amendment procedures.
For an organization, governance and succession mechanisms.
For a community, legitimacy may emerge through informal recognition and accepted social authority.
For an artificial identity, it might involve authorised memory updates, explicit constraints and traceable changes rather than arbitrary state injection.
This distinction matters because procedural legitimacy and moral legitimacy are not the same thing.
A transformation can be valid within a system while the system itself remains open to criticism.
Once that distinction is in place, the theory becomes much cleaner.
A continuity claim becomes stronger when four things can be defended.
First, there is an intelligible lineage between predecessor and successor.
Second, relevant identity-bearing structure survives or is meaningfully transformed.
Third, the transformation occurs through a legitimate process of change appropriate to the kind of entity involved.
Fourth, recognition of continuity is accountable to those conditions rather than resting on resemblance, branding or assertion alone.
This does not reduce continuity to a score.
It does something more useful.
It turns continuity into an argument that can be inspected.
Instead of asking only:
Is this the same thing?
we can ask:
What is being claimed to continue?
What made the predecessor recognisable as itself?
Which of those elements survived?
Which changed?
Which ended?
How did the successor arise from the predecessor?
What governed the transition?
What evidence supports the lineage?
What would falsify the claim?
And finally:
Did the surviving and transformed elements preserve the grounds on which recognition of continuity was warranted?
That question is harder to satisfy than resemblance.
Harder to fake than branding.
And considerably more useful than counting preserved files.
It also exposes a further complication.
Continuity may not always imply uniqueness.
Imagine a software project that forks.
Both branches share the same history until the split.
Both inherit substantial identity-bearing structure.
Neither appears from nowhere.
Which one is the continuation?
Possibly both.
Institutions, traditions, communities and political movements already behave this way. They branch. Successor groups form. Rival lineages emerge. Multiple entities can inherit meaningful relationships to the same predecessor.
So continuity and numerical identity should not be treated as identical questions.
One predecessor may produce a continuation, a fork, a descendant, a reconstruction, a replica, an archive or a replacement.
These relations are not interchangeable.
That means the more useful question may not be:
Is this the same thing?
but:
How is this related to what came before?
This turns continuity from a binary judgment into something closer to genealogy.
Artificial systems make that problem operational very quickly.
Modern language-model agents often separate the underlying model from memory, system instructions, retrieval systems, tools and external state. The model may change while other parts remain. Memory may exist outside the model entirely. Identity instructions may live somewhere else again.
Research on LLM agent memory already treats persistent memory as a distinct architectural component necessary for longer-term interaction. Work such as Generative Agents demonstrates systems that record experience, derive reflections and retrieve those memories later to guide behaviour.
But a system that remembers a history is not automatically the same system that produced it.
A copy can inherit the memory.
A fork can inherit the same instructions.
A replacement model can reproduce similar behaviour.
Two systems can begin from identical state and then diverge.
That makes the continuity problem explicit.
What carries identity?
What establishes lineage?
What authorises an update?
Can two forks both be successors?
Which inherited states remain descriptive history and which retain governing authority?
These questions should not be answered by casually importing theories of human identity.
Artificial agents are different objects.
But the framework gives us a disciplined way to ask the questions without pretending the answers are already known.
That distinction between what is established and what is proposed matters throughout this argument.
Existing research strongly supports several parts of the terrain.
Material persistence alone does not settle identity through time.
Organizations can remain recognisable through substantial adaptation.
Groups understand themselves through historical and cultural continuity.
Online communities can migrate between platforms, while migration can also fragment them.
Archives depend on provenance and context.
Preserving cultural artefacts is not equivalent to preserving living practice.
Institutional development is shaped by history, timing and sequence.
Those are established findings or well-developed theoretical positions.
The remaining moves are ours.
This essay proposes identity-bearing structure as a way of naming whatever actually matters to continuity for a given type of entity.
It proposes that continuity requires both identity-bearing structure and an intelligible lineage.
It distinguishes continuity from replication, reconstruction and descent.
It treats recognition as evidence rather than foundation.
It argues that inheritance should not automatically carry authority.
And it brings those propositions together in one central claim:
Continuity is a justified claim of succession.
That is the theory.
It should be tested as one.
The question that began this problem was simple:
What had to survive for participants to recognise this as the same ongoing thing?
That remains useful.
But recognition alone is too weak.
The stronger question is:
What had to survive for that recognition to remain warranted?
That change in wording carries the whole argument.
It moves us from appearance to structure, from preservation to history, from similarity to succession.
A thing may lose most of its visible form and continue.
Another may preserve almost everything visible and become something else.
What matters is not whether change occurred.
Change is inevitable.
What matters is whether the successor can account for the relation.
It should be able to say:
This is what made the earlier state what it was. This is what survived. This is what changed. This is what ended. This is how the transformation occurred. This is why the resulting state still belongs to the same thread.
That is a much stronger continuity claim than:
It looks the same.
Or:
We kept the files.
Or:
We kept the name.
The deeper test is whether what mattered remained legible across the crossing.
The room can disappear. The name can change. The machinery can be replaced. Membership can turn over. Language can evolve. Rules can be refused.
And continuity may still survive.
Not because nothing changed.
But because the change can still be understood as part of the history of the thing that changed.
That gives us the simplest form of the theory:
This changed, and we can explain why it is still this.
Sources and further reading
Ashforth, B. E. & Mael, F. “Social Identity Theory and the Organization.” Academy of Management Review, 14(1), 1989.
Fiesler, C. & Dym, B. “Moving Across Lands: Online Platform Migration in Fandom Communities.” Proceedings of the ACM on Human-Computer Interaction, 4(CSCW1), 2020. DOI: 10.1145/3392847.
Gioia, D. A., Schultz, M. & Corley, K. G. “Organizational Identity, Image, and Adaptive Instability.” Academy of Management Review, 25(1), 2000.
Mahoney, J. & Thelen, K., eds. Explaining Institutional Change: Ambiguity, Agency, and Power. Cambridge University Press, 2010.
Pierson, P. “Increasing Returns, Path Dependence, and the Study of Politics.” American Political Science Review, 94(2), 2000.
Ravasi, D. & Schultz, M. “Responding to Organizational Identity Threats: Exploring the Role of Organizational Culture.” Academy of Management Journal, 49(3), 2006.
Sani, F. et al. “Perceived Collective Continuity: Seeing Groups as Entities that Move Through Time.” European Journal of Social Psychology, 2007.
Society of American Archivists. Dictionary of Archives Terminology: Provenance.
International Council on Archives. Records in Contexts.
UNESCO. Convention for the Safeguarding of the Intangible Cultural Heritage, 2003.
World Wide Web Consortium. PROV-O: The PROV Ontology, W3C Recommendation, 2013.
Park, J. S. et al. “Generative Agents: Interactive Simulacra of Human Behavior.” UIST, 2023.
Citation note
This essay deliberately distinguishes established scholarship from the framework proposed here. Identity-bearing structure, continuity as a justified claim of succession, and the combined use of lineage, identity-bearing structure, legitimate transformation, and accountable recognition are proposed analytical concepts rather than claims of settled academic consensus.