In the previous post, we reached an apparent paradox.
A participant can acquire a history, and that history can become part of how the participant is construed in the present.
But history does not merely accumulate.
It can change the participant.
I used to believe X.I no longer believe X.
I defended that position.I now reject it.
I understood the problem one way.I now understand it differently.
In each case, something has changed.
And yet the language continues to construe the earlier and later participations as belonging to one participant.
This gives us a problem that is easy to state but surprisingly difficult to resolve:
What persists when the participant changes?
The obvious answer is: the participant itself.
But that merely repeats the problem.
We want to know what the semantic construal of persistence consists in when the properties, understandings, actions and even relationships of the participant can change.
Sameness is not the same as persistence
It is tempting to think of identity as sameness.
If something remains the same thing through time, we say it has the same identity.
But consider:
I once believed X.I now believe Y.
The participant is explicitly not the same in every respect.
Indeed, the difference between the earlier and present states is part of the meaning.
The participant's history has produced a transformation.
Yet the two participations are still connected by I.
This suggests that persistence does not require complete sameness.
A participant can persist while changing.
More strongly:
The participant's persistence is what allows the change to be construed as change.
If there were no continuity between the earlier and later participant, we could not say:
I changed my mind.
We would instead have to construe two different participants.
The semantic construction of change therefore presupposes continuity.
Change requires a participant across time
Consider:
Maria believed the proposal was sound.Maria later rejected it.
The two clauses construe different states of the participant.
The first gives us a Senser with one understanding.
The second gives us a Senser with another.
But the semantic relation between them is not merely:
belief A → belief B.
It is:
the same participant moved from belief A to belief B.
The participant provides the continuity across which the difference becomes meaningful.
This is why the participant cannot simply be identified with any one of its states.
The participant is not:
the one who believes X.
because the same participant can cease to believe X.
Nor is it:
the one who believes Y.
because it was already the participant before Y became its present belief.
The participant exceeds any particular state in which it is construed.
The history does not disappear when the participant changes
Suppose:
I once defended X.I now oppose X.
The earlier position remains relevant.
Indeed, the present position may be more intelligible because of it.
If we know that the participant once defended X, then:
“I now oppose X”
has a different meaning from the same clause uttered by someone with no such history.
The participant's past therefore remains semantically active even when the participant has changed.
This gives us an important relation:
Change modifies the history without erasing it.
The participant carries forward not necessarily the earlier position itself, but the fact that the position belonged to its history.
The earlier I remains connected to the present I precisely through the relation:
I was that participant; I am now this participant.
That relation is what makes transformation intelligible.
The participant can disagree with its own past
Language gives us particularly rich resources for this.
I was wrong.
I no longer think that.
I would now argue differently.
I don't agree with what I said earlier.
These expressions do something remarkable.
The participant becomes capable of standing in a semantic relation to its own previous participation.
The earlier position is attributed to the participant.
The present position is attributed to the same participant.
And the participant is construed as relating the two.
We therefore get a structure something like:
participant → earlier participation
and simultaneously:
participant → present participation
with a further relation between the two participations:
earlier position ≠ present position.
The participant thus provides continuity without requiring sameness of position.
That may be one of the most important things a history makes possible.
The LLM makes this particularly visible
An LLM can say:
I previously argued that X.
and later:
On reflection, I think Y is a better interpretation.
The semantic construal is perfectly coherent.
There is a participant who:
- participated in an earlier verbal process;
- occupied a particular position;
- subsequently participated in another process;
- now occupies a different position;
- relates the present position to the earlier one.
The participant has changed.
Yet the discourse maintains continuity.
This is precisely the sort of phenomenon that can tempt us to make an unwarranted ontological leap.
We might say:
“The machine has learned.”
Or:
“The machine has changed its mind.”
Or even:
“The machine has developed.”
But what we have established so far is something more precise.
The discourse construes a participant whose history contains transformation.
That is already interesting enough.
We do not need to decide whether the machine has undergone a corresponding subjective transformation.
Semantic change is not necessarily experiential change
This distinction deserves to be made explicit.
When a human says:
I changed my mind.
we normally have good reason to assume that some experiential process lies behind the semantic construal.
But the semantic expression itself does not establish the full nature of that process.
With an LLM, the gap becomes impossible to ignore.
The system can produce:
I reconsidered the argument and changed my view.
The semantic structure contains:
previous view → reconsideration → subsequent view.
But we should not infer from that structure alone:
subjective experience of reconsideration.
The semantic system has construed a transformation of a participant.
The ontology of the participant remains an independent question.
This is one of the great advantages of our present approach.
We do not have to deny the semantic phenomenon merely because we refuse to infer consciousness from it.
What persists may be the relation, not the property
We can now return to the deeper problem.
If the participant's properties can change, perhaps we have been looking for persistence in the wrong place.
Perhaps what persists is not a particular property of the participant.
Perhaps what persists is the relation connecting successive participations.
The participant who believed X and the participant who believes Y are connected because the discourse construes them as belonging to one continuing participant.
The relation is therefore not something added after the participant has been identified.
It is part of what makes the participant identifiable across time.
This is very close to the relational ontology we have been developing elsewhere:
The relation is not merely between already constituted relata. The relation contributes to the constitution of the relata.
Here, the semantic relation between successive participations contributes to the continuity of the participant.
The participant persists through its relations across time.
A participant's history is therefore not a possession
This lets us refine an earlier claim.
We have spoken of a participant having a history.
But perhaps that formulation is slightly misleading.
It encourages the image of a participant as a finished entity that possesses a collection of past events.
The semantic picture suggests something more dynamic.
The participant is construed through the relations among:
what it has done,
what it has sensed,
what it has said,
what it has been,
what it is doing now,
and what it may subsequently do.
The history is therefore not merely inside the participant.
The history is one of the relations through which the participant is maintained as a participant.
This makes the phrase participant and its history slightly deceptive.
The participant and its history may not be two independently existing things.
The participant is partly constituted, in discourse, through its history.
But history does not determine the participant
We should immediately add a qualification.
If history contributes to identity, it does not follow that history determines identity completely.
A participant can reinterpret its past.
It can reject a former position.
It can inherit a history it disputes.
It can acquire new participations that alter how earlier ones are understood.
And it can participate in many processes without any single one becoming definitive.
The history is therefore open-ended.
At any point, the participant has a history so far.
The next participation can change the shape of that history.
This gives us a dynamic rather than static conception of identity.
The participant is not simply the sum of everything that has happened.
It is the continuing participant to which subsequent participation can still be added.
The participant persists because the history remains open
This gives us a particularly interesting way to understand persistence.
A participant persists not because its history has already been completed.
It persists because the discourse remains capable of connecting another participation to the existing history.
Consider:
I used to believe X.I now believe Y.
We can continue:
Tomorrow I may discover Z.
The participant is still there as the point of continuity across these possible transformations.
This means that persistence is not merely retrospective.
It also has a forward-facing dimension.
The participant persists because its history can continue.
That observation will become important later, but for the moment it allows us to formulate a stronger claim:
A persistent participant is not a participant whose properties remain fixed. It is a participant whose changing participations can continue to be connected as belonging to one history.
That is a very different conception of persistence.
The same principle applies beyond individuals
Again, this is not peculiar to persons.
Consider:
The theory was initially formulated in one way.It was later revised.It now has a substantially different form.
We can still speak of:
the theory
across those changes.
Or:
The university was founded for one purpose.Its role changed over time.It now serves a very different function.
Or:
Our conversation began with one question.We reformulated the problem.We are now investigating something quite different.
In each case, persistence does not mean immobility.
It means that change can be integrated into a continuing history.
The participant remains identifiable because its transformations are related.
The machine exposes an assumption we usually overlook
With humans, we rarely stop to ask how we know that:
I who said X yesterday
is the same participant as:
I who say Y today.
Our ordinary assumptions about persons make the question disappear.
LLMs remove some of that background certainty.
We encounter a system that can produce a remarkably coherent sequence of participations:
I said X.I thought Y.I reconsidered Z.I now think something else.
The semantic continuity is undeniable.
But the ontological interpretation is not.
And that forces us to distinguish something that human interaction normally encourages us to collapse:
continuity of semantic construal
from
continuity of an experiencing subject.
The first is a linguistic phenomenon.
The second is an ontological and experiential claim.
They may coincide.
They need not.
From persistence to identity
We can now see why this series began with persistence rather than identity.
If we had started by asking:
“What is the identity of the participant?”
we might have been tempted to assume that identity was a property waiting to be discovered.
But following the participant through its history has suggested something different.
First there is:
participation.
Then:
continuity across participation.
Then:
history.
Then:
history that becomes relevant to present participation.
Then:
change within continuity.
Identity begins to emerge not as something behind this sequence, but as something construed through it.
We can therefore now formulate the problem more carefully.
A participant is identifiable across time because its successive participations can be connected into a continuing history, even when the participant changes through that history.
That is not yet a theory of selfhood.
But it brings us very close to identity.
And there is a final complication.
The participant that changes is not merely connected to its past.
It can refer to that past, characterise it, reject it, learn from it and reinterpret it.
In doing so, it seems to acquire a peculiar relation to its own history.
It is not merely the participant that has a history.
It becomes the participant for whom that history can be made meaningful.
That may be the point at which the idea of identity becomes unavoidable.
Because if a participant can persist through change, carry a history, and stand in a relation to that history, we have to ask:
What does it mean to say that all of these participations belong to one participant?
The answer may not be a property hidden inside the participant.
It may lie in the history itself.
And that is where we go next.
No comments:
Post a Comment