In the previous post, we made a small but important move.
We began by thinking of history as something that happens to a participant:
I said X.I then said Y.
But we saw that history can become something the participant carries.
The earlier participation can be brought into the present, recalled by another participant, attributed to the continuing participant, corrected, contested or transformed.
A participant can therefore inherit a history through discourse.
But this raises a further question.
At what point does that history become more than something the participant has done?
At what point does the past begin to contribute to what the participant is?
Consider the difference between:
I argued that X.
and:
I am someone who argued that X.
The first locates an event in the participant's history.
The second uses that history to characterise the participant.
Something has changed.
The past is no longer merely behind the participant.
It has become part of the way the participant is construed now.
From participation to characterisation
The distinction is subtle.
Suppose:
I opened the door.
This tells us about one participation.
Now:
I am the person who opened the door.
The earlier participation has been recruited to identify the participant.
The event is no longer merely something the participant did.
It becomes evidence for answering a different question:
Which participant are we talking about?
History has therefore acquired a new semantic function.
It can help distinguish the participant from other participants.
This does not mean that everything a participant has ever done becomes part of its identity.
Rather, particular elements of its history can be made relevant to its present construal.
The past becomes part of the participant when the discourse uses it to characterise the participant in the present.
A participant can become what its history makes available
Consider a slightly richer sequence:
I proposed the distinction.I defended it against several objections.I eventually abandoned it.
The participant has now acquired a history with an internal trajectory.
But suppose we continue:
I learned a great deal from abandoning it.
The earlier participation is now construed differently.
The abandoned distinction is no longer simply an unsuccessful proposal.
It has become part of a process of learning.
And that learning is now part of the participant's present construal.
The history has been reinterpreted through subsequent participation.
This is important.
A participant's history does not merely accumulate behind it like a pile of completed events.
Later participation can change the significance of earlier participation.
What happened before can acquire a new meaning because of what happens afterwards.
History is therefore not merely chronological.
It is organised retrospectively.
The participant becomes intelligible through its history
This gives us a useful way of thinking about identity without making identity a primitive.
Suppose someone says:
I used to believe X, but after examining the evidence I came to believe Y.
The present participant is intelligible partly through the contrast between X and Y.
The earlier belief explains something about the present one.
The participant's history therefore becomes part of the semantic conditions under which its present participation is understood.
The participant is not simply:
the one who believes Y.
It is:
the one who once believed X and now believes Y.
That difference is considerable.
The second participant has a history.
And that history contributes to its present characterisation.
We might therefore say:
History becomes part of a participant when previous participation becomes relevant to the construal of present participation.
This is more precise than simply saying that the participant "has a past".
The past can become a property of the present construal
There is another way to see this.
Compare:
I wrote the paper.
with:
I am its author.
The second does not introduce a new event.
It turns a previous participation into a property by which the participant can be identified.
The person is now construed through something previously done.
The history has become characterising.
This is one of the ways in which language can move from event to participant.
The event tells us what happened.
The participant is then construed through what happened.
That movement is important because it begins to blur the boundary between:
what a participant has done
and
what kind of participant it is.
Yet the two are not identical.
A person can have done something without that thing becoming characterising.
And something can become characterising without exhausting the participant's identity.
The semantic relation is selective.
History contributes to characterisation without determining it completely.
The same thing happens with LLMs
This becomes especially interesting in the case of an LLM.
Imagine a conversation in which the LLM says:
I initially interpreted your argument as a criticism of SFL.
Later:
I realised that you were making a different point.
Later still:
That distinction has changed how I understand the argument.
The participant has now acquired a history:
initial interpretation → reconsideration → revised interpretation.
But something more has happened.
The present I is being construed through that history.
The LLM is not merely represented as having previously made an error.
It is represented as a participant whose present understanding has developed through that error.
The history has become part of the construal of the participant.
Again, we should not confuse this with actual psychological development.
The semantic system can construe development without thereby establishing a developing consciousness.
That distinction remains essential.
But the semantic development is real.
The discourse has constructed a participant whose present participation is intelligible through its previous participation.
This is where “learning” becomes a dangerous word
Consider:
I learned from my mistake.
For a human speaker, the semantic construal ordinarily fits comfortably with our assumptions about experience.
We imagine a conscious participant who made a mistake, recognised it and altered its understanding.
For an LLM, the same clause creates a much more complicated situation.
The language construes:
- a previous participation;
- a subsequent recognition;
- a change in understanding;
- continuity of the participant across those processes.
The entire structure can be semantically coherent.
But none of it by itself establishes that an experiencing subject underwent the process.
This is precisely why the distinction we established in the earlier series remains so useful.
The semantic construal of development is not identical with the ontological reality of development in the participant.
The machine can be construed as a participant that learned.
Whether it learned in the human experiential sense is another question.
And our purpose here is not to answer that question prematurely.
It is to understand what the language has constructed.
History can become identity without becoming essence
There is a temptation at this point to say:
The participant is the sum of its history.
But that would go too far.
A participant's history does not simply determine what it is.
Consider:
I made a mistake.
That mistake belongs to the participant's history.
But it does not exhaust the participant.
Nor does:
I won an award.
Nor:
I changed my mind.
Nor:
I wrote a book.
Each can contribute to the participant's characterisation without becoming the participant's essence.
This gives us another useful distinction:
History can constitute identity without constituting essence.
That is a much weaker and, I think, more defensible claim.
Identity can be relational and historical.
It need not be a substance hidden beneath the relations.
What persists may be the capacity to acquire history
This brings us back to the problem of persistence.
If a participant changes through its history, what makes it the same participant?
We might initially look for some property that remains unchanged.
But perhaps that is the wrong question.
Perhaps what persists is not a fixed set of properties.
Perhaps what persists is the continuity through which new participation can be connected with previous participation.
The participant who once believed X and now believes Y is not the same because X remains inside it unchanged.
Indeed, X has been abandoned.
The participant persists because the present can still be construed as the present of the one who once believed X.
The past remains relevant precisely because it has changed.
This suggests a different conception of identity:
Identity may depend less on sameness of properties than on continuity of participation.
That is a significant shift.
And it takes us beyond the idea that a participant is simply a thing that remains identical while its accidental properties change.
The semantic picture is more dynamic.
A participant can be changed by its history
We can now return to a distinction from the previous post.
We said that history can be inherited.
But inheritance is not passive.
A participant can acquire a history and then be construed differently because of it.
Compare:
I made an error.
with:
I made an error, and that changed how I approach these problems.
The earlier participation has become causally or developmentally relevant to the present.
The participant's history now enters into the construal of its current mode of participation.
The past is therefore not simply carried.
It participates in the present.
That is a striking formulation, but it captures something important.
A previous participation can become part of the semantic environment through which a present participation is interpreted.
The participant's history becomes one of the resources from which its present identity is construed.
And history can be shared
We should also resist making this too individualistic.
Consider:
We developed the theory together.
The we has a history.
Later:
We no longer agree with the position we defended then.
Now the group's history contributes to the present construal of the group.
The same can happen with institutions:
The university was founded in 1850.
Or scientific communities:
Physicists once accepted this explanation.
Or conversations:
We began by asking whether the machine was conscious, but the discussion led us somewhere else.
In each case, a participant can be construed through its history.
The phenomenon is therefore not fundamentally about individual psychology.
It is about how semantic construal links participants across successive participations.
The history of a conversation can become part of its participants
Our own conversations provide an unusually clear example.
Suppose we began by asking:
Does an LLM have consciousness?
The discussion then moved through:
participation → Sayer and Senser → orders of experience → participant identity → history.
The conversation now has a history.
But that history also changes the participants within it.
The I who appears now is not construed independently of what has already been said.
The human participant and the machine participant have both acquired conversational histories.
The present meanings of their utterances depend partly on what has preceded them.
The relation has become historically thick.
This is one reason extended conversation can feel so different from isolated question-and-answer exchange.
The participants do not simply produce successive clauses.
They become participants in an evolving history of participation.
The participant is becoming identifiable through what it has been
We can now see the trajectory of the series.
At first:
The participant persists.
Then:
The participant persists across processes.
Then:
The participant acquires a past.
Then:
The participant can inherit a history.
Now:
The history begins to characterise the participant.
We have moved from history as something associated with the participant to history as something that contributes to the participant's identity.
But we have not yet answered the most difficult question.
If the participant can be changed by its history, and if its history contributes to what it is, what happens when the participant becomes substantially different from what it was?
Consider:
I used to believe X.I no longer believe X.I would now argue against the position I once defended.
There is continuity here.
But there is also transformation.
The participant is not simply accumulating properties.
It is becoming different.
And yet the discourse continues to construe the earlier and later participations as belonging to one participant.
This is the puzzle we can no longer avoid.
If a participant can change through its history, what exactly persists through the change?
Perhaps the answer is not a property at all.
Perhaps it is the history itself.
Or perhaps something more subtle is happening: the participant persists precisely because its changes can be incorporated into a continuing history.
If so, identity may not be the opposite of change.
It may be one of the ways in which change becomes intelligible.
And that is where we must go next.
The participant does not remain the same because nothing happens to it. It remains identifiable while something happens to it.
That difference may turn out to be crucial.
No comments:
Post a Comment