Assessment Through Williamsonian and Finean lenses: Brandom (1)
What Brandom is trying to do sits slightly to one side of the Fine–Williamson dispute, but it intersects with it in a revealing way. The quickest orientation is this: Brandom shifts the focus from truth conditions to inferential role. Instead of asking when two sentences are true in the same situations, he asks what follows from them, what they entitle you to say, and what commitments they bring with them. Meaning is use, but in a very specific sense, use as governed by norms of inference (Brandom 1994).
Brandom is not primarily operating at the extensional level, because he is not concerned simply with what sentences are true of. Nor is he straightforwardly intensional in the Williamson sense, because he is not starting from possible worlds and modal profiles. Nor is he straightforwardly hyperintensional in Fine’s sense, because his distinctions are not grounded in truthmakers or metaphysical dependence. Instead, he is offering a pragmatic, normative reconstruction of logic, where the basic units are commitments and entitlements within a social practice.
Now, how would Williamson react to this? Williamson would be sympathetic to the idea that inference matters, but he would resist Brandom’s attempt to make inferential role fundamental. For Williamson, logic is about objective consequence relations, not about the practices of speakers. When we say that one claim follows from another, that relation is not constituted by what people are prepared to infer, or by the norms governing their discourse. It is a fact about how things are, or about the structure of propositions, which our practices aim to track but do not determine.
This is why Williamson rejects what he sees as the “linguistic or conceptual turn” in philosophy, where meaning is analysed in terms of use or practice. He would say that Brandom risks projecting features of our inferential behaviour onto the world itself. The danger is similar to the one he identifies in hyperintensional arguments. We notice that certain inferences feel good or bad, that some explanations are more satisfying than others, and we build those judgments into our account of meaning. But those judgments may be guided by heuristics, by training, by communicative convenience. They are not a secure foundation for logic. From Williamson’s perspective, Brandom’s inferentialism is therefore too epistemic and practice bound. It ties meaning to what agents can do, justify, or recognise. Williamson’s own position insists that truth, necessity, and consequence outrun those capacities. We can fail to see that one claim follows from another, but that does not mean the consequence relation is not there. So he would likely treat Brandom’s framework as, at best, a theory of how we handle logic, not a theory of what logic is.
Now consider Fine. Fine is much closer to Brandom in spirit, though still importantly different. He shares the sense that intensional frameworks miss something, that there are distinctions finer than truth across possible worlds. Where Williamson says “necessary equivalence is enough,” Fine says “no, there are differences in structure, dependence, and explanation.” Brandom, in his own way, is also dissatisfied with a purely truth conditional picture. He wants to capture how claims are connected through inference, how they support and constrain one another. So Fine would likely see Brandom as identifying a real phenomenon. Inferential relations do seem to matter. The meaning of a concept is tied to what follows from it and what would count as a reason for it. That resonates with Fine’s own interest in grounding and dependence, where some truths hold in virtue of others.
However, Fine would resist Brandom’s attempt to reduce everything to inferential role. For Fine, the key relations are not just inferential but metaphysical. When one fact grounds another, or when one truth makes another true, that is not merely a matter of what we are committed to infer. It is a matter of how reality itself is structured, or at least how our best semantics should represent that structure. So while Brandom focuses on norms of reasoning, Fine wants to preserve a distinction between those norms and the underlying relations they aim to track.
You can put the contrast like this.
Brandom says:
meaning is constituted by inferential role within a practice.
Fine says:
inferential patterns reflect deeper relations of dependence, which are not exhausted by practice.
Williamson says:
inferential patterns should be explained by objective truth and consequence relations, not the other way around.
That triangulates the positions quite clearly.Now bring this into philosophy of education, where the stakes become more concrete.Brandom’s view is immediately attractive in educational terms. Learning becomes a matter of entering a space of reasons, of taking on commitments, of being able to justify claims and draw appropriate inferences. To understand a concept is to know what follows from it and what counts as a reason for it. This aligns closely with pedagogical practices that emphasise explanation, argument, dialogue, and justification. A student shows understanding by participating competently in these inferential practices.
Williamson would accept much of this at the level of pedagogy, but he would resist taking it as an account of knowledge itself. A student might be able to give the right inferences, to play the inferential game well, and yet still lack knowledge if their beliefs are not appropriately connected to truth. Conversely, a student might know something even if they are not yet able to articulate all its inferential consequences. So for Williamson, inferential ability is evidence of knowledge, but not its essence.
Fine would be interested in how inferential roles relate to underlying structures. Two students might draw the same inferences, but differ in how their understanding is grounded. One might rely on memorised patterns of inference, the other on a grasp of the principles that make those inferences valid. This is a hyperintensional distinction. It is not visible at the level of which inferences are endorsed, but at the level of how those inferences are supported. Fine would want an account that can capture that difference, something Brandom’s framework risks flattening if it focuses only on overt inferential commitments.
There is also a deeper tension here that connects to my broader concerns about AI and education. Brandom’s inferentialism fits very naturally with systems that can reproduce inferential patterns. If a system can take in claims and generate appropriate consequences, give reasons, respond to objections, then on a purely inferentialist view it begins to look as though it participates in the space of reasons. That is precisely the sort of phenomenon we are now encountering. Williamson would push back by insisting that matching inferential behaviour is not enough. The question is whether the system’s outputs track truth in the right way across possibilities. Fine would push further by asking what grounds those outputs, what makes them correct, and whether the system has the relevant internal structure. Brandom’s framework, if taken on its own, risks treating successful participation in inferential practice as sufficient for understanding, which is exactly where many current debates about AI become most contentious.
So the overall picture is this. Williamson would see Brandom as over intellectualising practice and under theorising truth. Fine would see him as identifying important structural features but mislocating them at the level of practice rather than grounding. Brandom himself is trying to reorient the entire discussion, to say that meaning and logic are not primarily about mapping language onto a pre given world, but about the norms governing our reasoning.
For philosophy of education, the encounter is productive. It shows three different ways of understanding learning. One focuses on truth and modal robustness, one on grounding and structure, and one on participation in inferential practice. Each captures something real. The difficulty is that none of them, on its own, seems sufficient to account for what we actually care about when we say that someone has learned something.
So, if one takes seriously the differences between Williamson, Fine, and Brandom, then one does not simply get three abstract theories of meaning or logic. One gets three rather different pictures of what it would be to show that one has learned something, and therefore three different tendencies in assessment. The differences are not superficial. They arise from what each philosopher takes to be most basic, what counts as the same piece of knowledge, and what sort of evidence could show that a learner possesses it.
The basic shape of Williamson's position is that meaning, thought, and knowledge should not be tied too closely to the accidents of wording, presentation, or immediate human accessibility. His broader philosophical instinct is realist and anti-psychologistic. By this I mean that he does not think truth, logical consequence, or modality depend upon what we happen to find easy to grasp, explain, or verify. Things can be true even when we cannot prove them, and one claim can follow from another even when we do not notice that it does. In the debate with Fine, this emerges in his defence of an intensional framework. (Very roughly, two propositions are the same if they have the same truth conditions across possible worlds, that is, if they are true in exactly the same possible situations.) The point of possible worlds talk is not to indulge fantasy. It is a disciplined way of asking what would remain the same across different cases, variations, and circumstances.
Once one translates this into education, a fairly clear assessment tendency follows. A Williamsonian assessment would be suspicious of overvaluing the exact route by which a student arrived at an answer, or the local stylistic features of an explanation, if those features do not mark a difference in the underlying knowledge state. What should matter is whether the student’s grasp tracks the truth robustly across relevant variations. The idea of relevant variation is important. If a student really understands a mathematical rule, a historical concept, or a scientific principle, then that understanding should not be confined to one familiar wording or one rehearsed example. It should survive changes in surface form. So the appropriate assessment, from this point of view, is one that tests modal robustness. That phrase can sound technical, but its meaning is simple enough. One asks whether the student would still get things right if the presentation changed, if the case were slightly altered, if an example were novel, if a distracting but irrelevant feature were introduced, or if the same concept appeared in a different domain.
This would favour assessments built around families of cases rather than single performances. One would want to know not merely whether the student answered this question correctly, but whether she can identify the same structure when the problem is rephrased, disguised, transferred, or set under pressure. A Williamsonian teacher would therefore value unseen problems, counterexample testing, transfer tasks, and questions that vary context while preserving conceptual structure. The educational rationale would be that genuine knowledge is not exhausted by success in the actual case, which would be merely extensional. Nor is it best captured by a richly narrated account of one’s own thought process, which may reveal as much about fluency or confidence as about knowledge. Instead, one looks for a pattern of correctness across possible variations of the task.
There would also be a corresponding caution about certain current assessment fashions. Reflection diaries, open ended self reports about learning, and highly personalised accounts of understanding might all be treated with some reserve. Not because they are worthless, but because Williamson’s critique of projection would make him suspicious of the slide from how things are said to what is known. In the debate, he worries that philosophers often take differences in explanatory usefulness or wording and project them onto reality itself. In education, the analogue would be a concern that teachers project differences in eloquence, confidence, or pedagogic helpfulness onto differences in knowledge. A student may offer a beautifully phrased explanation because she has excellent verbal habits, while another may know equally well but speak more awkwardly. So a Williamsonian assessment regime would try to strip away some of these misleading cues and ask instead whether the learner can reliably discriminate truth from error across a structured range of cases.
This does not mean that explanation would disappear. Williamson is not against explanation. But he would likely treat explanation as evidence of knowledge rather than as the essence of knowledge itself. An assessment designed in his spirit would therefore use explanation diagnostically. It would not ask, in a romantic way, for the student’s voice as such. Rather, it would ask for an explanation because the explanation can reveal whether the student’s grasp generalises, whether she can cope with variant cases, and whether she is responsive to logical relations. The gold standard would remain something like stable cognitive contact with objective structure.
The result would probably be an assessment system that looks, on the surface, more conventional than Fine’s or Brandom’s, though ideally more intelligent than much conventional testing. It would not be satisfied by rote recall, because rote recall may fail under variation. Nor would it be impressed by one off performances of apparent depth if they do not carry across cases. Its preferred format might be cumulative, synoptic, comparative, and deliberately varied. One can imagine mathematics papers with the same principle tested under many surface transformations, science assessments requiring application of a law in unfamiliar settings, or literary assessments that test whether a student can recognise a theme or formal relation in texts not previously rehearsed. The philosophical principle is that sameness of knowledge is tracked intensionally, by what stays true across relevant possibilities.
Fine’s position would yield a rather different sensibility. He does not reject the importance of modal variation, but he thinks the intensional test is too coarse. Necessary equivalence is not enough for sameness. Two claims can be true in all the same possible worlds and yet differ in structure, dependence, explanation, or what makes them true. His favourite kind of example is logically simple but educationally fertile. “P” and “P or P and Q” are true in exactly the same situations, yet Fine insists that they should not be treated as identical, because the second statement carries extra structure. In his truthmaker framework, the two differ in what could make them true. A truthmaker is whatever in the world, or in a semantic model of the world, does the work of making the statement true. Even if two statements never differ in truth value, they may differ in the range or organisation of their truthmakers. That is the hyperintensional thought.
Translate this carefully into learning and the consequences are significant. Fine would not be content with an assessment that merely establishes that a student gets the right answer across multiple cases. Two students may be intensionally equivalent in that sense and still differ hyperintensionally. One may understand a theorem by seeing its proof and its place in a network of concepts, while another may have learned a reliable shortcut that works across all the tasks set so far. One may be able to reconstruct the result from underlying principles, while another merely recognises the pattern. Their outputs coincide, perhaps even across a generous range of possible questions, but the inner structure of their understanding differs. The first student’s answer is made true, in educational terms, by a richer and more articulated grasp. The second student’s answer is made true by a thinner route. Fine’s assessments would therefore be designed to reveal not only whether the student can succeed, but what makes that success possible. One can hear the truthmaker language becoming educational without much strain. What in the learner’s understanding is doing the truthmaking work here? Is the answer supported by memorised association, by inferential reconstruction, by grasp of dependence relations, by sensitivity to what follows from what, by awareness of why this claim holds in virtue of those other claims? Fine’s metaphysical interest in grounding would matter here. Grounding is the relation in virtue of which one fact holds because of another, not merely alongside it. If educationally translated, grounding becomes the relation in virtue of which one piece of understanding rests on others.
An assessment inspired by this would need to be much more diagnostic of internal structure. It would ask not only for the answer and not only for an explanation, but for the dependence pattern of the explanation. Show how this conclusion follows from these premises. Show which steps are essential and which are dispensable. Reconstruct the argument in another order. State what the result depends on, and what it does not depend on. Identify which proposition is doing the real work. Distinguish the reason from a merely accompanying fact. In literature, this might mean not just offering an interpretation but identifying which textual details ground it and which are ornamental. In science, it might mean not just stating a law but showing which features of the setup are explanatorily relevant. In history, it might mean distinguishing causal conditions from background conditions, triggers from structures, and justification from chronology.
Fine’s concern with one to many relations between language and reality would also shape assessment. In the debate with Williamson he argues that a sentence may correspond to reality in multiple ways. Educationally, this suggests that a correct response may be underwritten by multiple structures of understanding. A good assessment would try to discriminate among these rather than flatten them into a single mark. Viva style questioning, oral defences, proof reconstruction, layered “why” questions, and tasks that force students to distinguish between equivalent formulations would all fit naturally here. One might, for example, ask whether the student understands why two formulas that are extensionally or intensionally equivalent are nevertheless not conceptually identical in role. Or ask the student to compare two solutions that reach the same answer and explain whether they are equivalent, more perspicuous, more general, or more dependent on special assumptions.
Such assessments would likely resist the standardisation that large institutions often prefer. Fine’s framework is not hostile to rigour, quite the opposite, but its rigour is finer grained. It seeks to expose structure, not merely count successful outputs. That would make the resulting assessment ecology more qualitative in one sense, though not soft. It would be exacting about what depends on what. A student who arrives at the right answer by an accidentally reliable route would not count as equal to one who sees the internal articulation of the topic. From within Fine’s technical machinery, this is because the two students are not hyperintensionally equivalent. Their performances coincide extensionally and perhaps intensionally, but not in the finer structure that individuates genuine understanding.
One can now see why Fine’s perspective is especially suggestive in the age of generative AI. If assessment remains merely extensional, or even broadly intensional in a limited way, then a machine or a student using a machine can produce outputs that look indistinguishable from understanding. Fine’s framework gives philosophical language for resisting that flattening. The issue is not only whether the output is correct, or even whether it remains correct across a range of tasks. The issue is what grounds the correctness. If the answer is generated by statistical pattern completion rather than by the student’s own grasp of dependence relations, then the educational event is not the same, even if the visible performance is.
Brandom would push assessment in a third direction. His inferentialism begins from the thought that the content of a claim is determined not primarily by what it pictures or by what worlds make it true, but by its place in a network of commitments and entitlements. To understand a concept is to know, practically and normatively, what follows from applying it and what would count as a reason for applying it. Meaning is not exhausted by truth conditions. It is given by inferential role. Brandom’s philosophy is therefore social and normative in a way that neither Williamson’s nor Fine’s is. He asks us to see language users as participants in a game of giving and asking for reasons. A claim has content because of the moves it licenses and forbids in that game. One can put this very simply. Williamson asks whether the student’s claims track truth across possibilities. Fine asks what structure makes those claims true. Brandom asks whether the student can use the claims properly in reasoning, justification, and challenge. The educational consequences are immediate. A Brandomian assessment would care above all about participation in the space of reasons. Can the learner undertake a commitment, defend it when challenged, infer appropriately from it, recognise what else it commits her to, and revise her position in the light of objections? Learning would not be measured chiefly by possession of isolated true beliefs, nor even by internal grounding alone, but by normative fluency within a practice of reason giving.
This would favour dialogical and discursive assessments. Seminars, oral examinations, debates, structured peer critique, written argument with rebuttal, and tasks requiring students to justify, defend, revise, and extend claims would be especially apt. One can imagine an assessment in philosophy, history, or science where the student is not simply asked for the right answer, but is drawn into a sequence of inferential moves. If you say this, what follows? What would count against it? Which earlier commitment are you now withdrawing? Are these two claims compatible? What entitles you to make that assertion? Can you distinguish evidence from implication? The point is not merely to check whether the student has memorised propositions, but whether she can inhabit them as commitments within a rational practice.
Brandom would also make room for a social dimension of assessment often underplayed elsewhere. Because the space of reasons is public, assessment should not always be a private encounter between student and examiner. There is philosophical justification, on his view, for forms of peer assessment, collaborative inquiry, and dialogical testing, provided these are rigorously structured. A learner’s understanding is shown in part by her ability to answer to others, not just to produce polished monologues. This could justify more live, in person, real time assessment forms in which the student’s inferential responsiveness is visible. In the current context, that may have practical appeal as well, because such assessments are harder to outsource to AI.
Yet Brandom’s approach would bring its own risks. Williamson would warn that one can become too impressed by discursive fluency. A student may navigate the game of reasons competently without secure contact with truth. Fine would add that two students may perform the same inferential moves while differing in the deeper structure that supports those moves. One may genuinely see why a conclusion follows, while another has simply acquired the social knack of moving from one sentence to another in accepted ways. So Brandomian assessments could drift towards rewarding verbal agility, rhetorical confidence, or conformity to disciplinary discourse unless carefully disciplined.This is where the three positions become mutually corrective. A purely Williamsonian regime may underrate explanation, dialogue, and the learner’s own articulation of reasons. A purely Finean regime may become so interested in inner structure that it is difficult to implement fairly and publicly at scale. A purely Brandomian regime may reward inferential performance without sufficient concern for truth tracking or grounding. But if one speculates about ideal assessment forms from within each philosophy, the contrasts remain clear.
Williamson would design assessments that stress robustness under variation, conceptual transfer, error discrimination, and performance across a wide modal field of cases. Fine would design assessments that expose grounding structure, reveal what the student’s answer depends upon, and distinguish superficially equivalent performances by probing their inner articulation. Brandom would design assessments that place the learner in the public space of reasons, requiring her to undertake commitments, infer from them, answer challenges, and revise in a norm governed practice.
One can imagine how these differences might play out in a single concrete subject. Take literary interpretation. A Williamsonian assessment would examine whether the student can identify a theme, formal feature, or conceptual tension across unfamiliar texts and changing contexts, showing robust discrimination. A Finean assessment would ask what in the text makes the interpretation apt, which features ground the claim, whether two interpretations with the same overall conclusion differ in explanatory structure, and which details are truly doing the work. A Brandomian assessment would turn on the student’s ability to enter the disciplinary conversation, advance an interpretation, defend it with reasons, respond to counterreadings, and recognise the commitments entailed by her claims about genre, voice, or historical setting.
The same pattern extends to professional and practical education. In medicine, a Williamsonian assessment would test whether diagnostic judgement transfers across varied cases. A Finean one would probe whether the student sees which symptoms are causally and explanatorily central rather than merely correlated. A Brandomian one would assess whether the trainee can justify decisions, weigh reasons, answer challenge from colleagues, and participate responsibly in a norm governed professional practice.
At the deepest level, then, assessment follows ontology and semantics more closely than educational theorists often notice. If one thinks sameness is extensional, assessment will focus on actual outputs. If one thinks sameness is intensional, assessment will test robustness across possibilities. If one thinks sameness must be hyperintensionally fine grained, assessment will investigate structure, dependence, and grounds. If one thinks content is inferential role, assessment will put the learner into the game of giving and asking for reasons.
For contemporary education, especially under conditions of pervasive AI assistance, the pressure to clarify these distinctions is growing. Systems that reward extensional sameness, the same visible product, are now the easiest to game. Systems that test broad intensional robustness may fare better, though even there machine assistance can intrude. Hyperintensional and inferentialist assessments offer richer resources, because they ask what grounds an answer and how a learner inhabits a network of reasons. Yet they also demand more from institutions. They require time, expertise, and forms of judgement that cannot be reduced to automated marking.
The lesson is not that one philosopher has already solved the problem of assessment. It is rather that each reveals a different educational value. Williamson preserves the objectivity of knowledge and the importance of robust transfer. Fine preserves the inner structure of understanding and the difference between genuine grasp and merely coincident success. Brandom preserves the public, normative, reason giving character of learning as a practice. A serious assessment culture would need to decide not only how to measure performance, but which of these values it is willing to build into its very idea of what counts as learning.
References
Brandom, R. (1994). Making It Explicit. Cambridge, MA: Harvard University Press.
Brandom, R. (2000). Articulating Reasons. Cambridge, MA: Harvard University Press.
Fine, K. (2017). The Limits of Abstraction. Oxford: Oxford University Press.
Williamson, T. (2000). Knowledge and Its Limits. Oxford: Oxford University Press.
Williamson, T. (2013). Modal Logic as Metaphysics. Oxford: Oxford University Press.