ChatGPT Is Said in Many Ways

1. A Very Short Classroom Scene

Two teachers, A and B, were working together in team teaching when they asked their students to research a young person who had done something significant for humanity. When A asked how they would begin, one student said Google. Another said ChatGPT.

B suggested a third possibility: use ChatGPT to locate articles and sources, then go to those sources, read them, and evaluate them. It could combine the orientation provided by ChatGPT with direct access to sources on the web.

A immediately objected:

“No, no, no. You shouldn't use AI for searching. That is a mathematical prediction. It is inaccurate. It has no value.”

B did not argue. But the sentence stayed with her.

How can the same tool appear so differently to two people standing in the same classroom?

2. The Same Thing is Said in Many Ways

Aristotle gave us the famous formula “the good is said in many ways” in his critique of Plato’s Forms.

The point here is not Aristotle's theory of the Good. What matters is the structure of the expression. How can the same thing be said in many ways?

The question can be applied to ChatGPT. It can appear as a search tool, a statistical predictor, a writing assistant, a shortcut, a source-finding device, a threat to authorship, or a pedagogical resource.

These are not simply different meanings of the word. They are different ways of positioning the same object within a network of relations. What ChatGPT is taken to be depends, at least partly, on the terms and practices with which it is associated.

This gives us a different way of understanding the phrase: ChatGPT is said in many ways because it can enter different constellations.

3. The Constellation Around a Word

This is where Saussure becomes useful. For Saussure, the value of a sign is constituted by its relations within a system rather than by something it possesses in isolation.

A constellation, in this sense, is not necessarily a theory consciously held by an individual. It is a cluster of associated terms, contrasts, practices, and assumptions that gathers around a word and helps determine how it is understood.

Consider two possible constellations around ChatGPT.

Constellation A

ChatGPT
→ abundance
→ access
→ source discovery
→ reading
→ comparison
→ critical evaluation
→ student judgment
→ learning

Within this constellation, ChatGPT can appear as a way of navigating an abundance of information and directing students towards material they can then examine for themselves.

Constellation B

ChatGPT
→ mathematical prediction
→ automation
→ lack of human understanding
→ lack of effort
→ shortcut
→ substitution
→ unreliable knowledge

Here the same word occupies a very different position. It appears less as a means of orientation than as a possible substitute for human effort and judgment.

Saussure's distinction gives us here another way to glimpse the process: a term acquires value through what accompanies it and through what it contrasts with.

The word is the same. The constellation is different, and therefore its value is different.

4. The Paradox of “Mathematical”

Let us return to A's sentence: “That is a mathematical prediction. It is inaccurate. It has no value.”

The interesting question is not simply whether the statement is correct. It is: What does “mathematical” mean inside this constellation?

In many contexts, “mathematical” carries associations such as precision, objectivity, measurability, and reliability. Here, however, the movement is almost the reverse:

mathematical → mechanical → predictive → lacking understanding → unreliable.

The same term can therefore function as a mark of authority in one constellation and as a reason for distrust in another.

But there is also an epistemological leap in A's statement. “It is a mathematical prediction” does not by itself entail “it is inaccurate” or “it has no value.” To reach that conclusion, an additional premise is needed: if a system produces answers through mathematical prediction rather than understanding, those answers cannot count as reliable or valuable knowledge.

That premise is not mathematical. It is epistemological. It concerns what kinds of processes are allowed to produce knowledge that counts.

The argument sounds mathematical, but its conclusion rests on an epistemology.

5. Double Valuation

This is where the idea of double valuation becomes useful. The same feature can acquire a different status depending on the relations in which it appears.

What counts as knowledge depends on the conditions under which something can appear as meaningful knowledge in the first place. The distinctions that separate the reliable from the unreliable, the valid from the invalid, are themselves part of the system within which an utterance acquires its force.

We therefore do not need to ask simply whether AI is reliable.

The more fundamental question is: What makes one form of knowledge count as reliable here and another as unreliable?

The issue is not merely whether mathematical prediction is good or bad. It is how the same kind of predictive operation can acquire different values depending on the system in which it is placed.

6. Why Arguments About AI Become So Polarized

This helps explain why arguments about AI can become so polarized.

The question is often framed as: Is ChatGPT good or bad?

But people asking this question may be starting from very different constellations.

For one:

ChatGPT → abundance → access → sources → critical reading → judgment.

For another:

ChatGPT → automation → shortcut → lack of effort → substitution.

The two positions are not merely different answers to the same question. They organize the question differently. What appears as a useful instrument in one constellation can appear as a threat to human judgment in another.

This does not mean that every disagreement is merely verbal, or that all positions are equally valid. It means that some disagreements may begin earlier than we think.

Before asking why two people disagree about ChatGPT, it may be worth asking what each person has already made ChatGPT into.

They use the same word, but they may be saying it in different conceptual languages.

7. The Constellations We Cannot See

There is, however, a final difficulty. Once we learn to look for the constellation within which someone else's judgment becomes intelligible, it is tempting to assume that we can see what they cannot.

But what about our own?

B sees: ChatGPT → abundance → sources → judgment.

A sees: ChatGPT → prediction → automation → lack of effort.

Yet B's constellation may feel like simple common sense to her precisely because it is B's own.

This is perhaps the most difficult step in learning to think relationally. We can become very good at identifying the assumptions behind other people's views while leaving our own assumptions untouched.

The hardest systems to see are the ones that make our own way of seeing seem obvious.

ChatGPT is said in many ways. So are knowledge, authorship, learning, evidence, and human judgment. Sometimes a disagreement does not begin with two different answers. It begins with two different constellations around the same word.

Sometimes the disagreement begins before the argument even starts.

References

Aristotle. (1999). Nicomachean ethics (T. Irwin, Trans., 2nd ed.). Hackett Publishing Company.

Aristotle. (1995). The complete works of Aristotle: The revised Oxford translation (J. Barnes, Ed.). Princeton University Press.

Derrida, J. (1976). Of grammatology (G. C. Spivak, Trans.). Johns Hopkins University Press.

Foucault, M. (1971). L'ordre du discours: Leçon inaugurale au Collège de France prononcée le 2 décembre 1970. Gallimard.

Saussure, F. de, Bally, C., Sechehaye, A., & Riedlinger, A. (1983). Course in general linguistics (R. Harris, Trans.). Duckworth.