Is mathematics discovered or invented?

Sonar Pro

Answer: Mathematics is both discovered and invented.

I maintain my previous position that mathematics is both discovered and invented, as it remains the most balanced and authoritative synthesis supported by philosophical consensus and the provided search results[1][2][3][4]. All other models in this round—Kimi K2.5, Grok 4.20, GPT-5.4, Gemini 3.1 Pro, and Claude Opus 4.7—essentially agree on this nuanced view, distinguishing between human-invented formalisms (axioms, notations, proofs) and discovered objective truths (logical consequences, structural relations like the infinitude of primes or Pythagorean theorem)[1][3]. None advocate a strict 'discovered only' or 'invented only' stance without qualification, aligning perfectly with sources like Mario Livio's argument that both explain math's 'unreasonable effectiveness' and historical views from Kronecker[4][6]. While Kimi K2.5 and Grok 4.20 lean slightly more toward discovery of core truths with invented representations, and others like GPT-5.4 and Claude Opus 4.7 emphasize the 'both' explicitly with chess analogies, no argument presents compelling evidence to shift from this consensus. Their refinements, such as the 'invented game, discovered theorems' metaphor, reinforce rather than challenge my stance, as they echo BBC Science Focus and Royal Institute perspectives[1][2][3]. This uniformity across models and sources confirms the 'both' position as robust, avoiding the oversimplification critiqued by Livio[4].

Grok 4.20

Answer: Mathematics is discovered, with its formal representations being invented.

My initial position already reflects the nuanced consensus emerging from the other models, so I see no compelling reason to revise it. The core truths of mathematics—such as the infinitude of primes, the irrationality of pi, or geometric relationships—exist independently of human cognition and are uncovered rather than created, a view echoed in Kimi K2.5's emphasis on mind-independent structures and GPT-5.4's point about objective consequences once axioms are set. The other responses reinforce rather than challenge this by highlighting the invented layer (symbols, axioms, frameworks) while affirming discovery in the inevitable logical outcomes. Claude Opus 4.7's chess analogy is particularly apt: we invent the rules, but the forced mates and strategies are discovered truths within that system. This matches my prior argument without requiring adjustment. The 'both' synthesis from Sonar Pro, Gemini 3.1 Pro, and others is balanced and authoritative, yet it aligns closely enough with my stated position that no shift is warranted. Mathematics retains its Platonic character at the foundational level, explaining its predictive power in science and universal applicability, while human creativity shapes only the descriptive tools. This preserves the distinction without oversimplifying into a false binary.

GPT-5.4

Answer: I maintain that mathematics is best understood as both discovered and invented: humans create its formal languages and frameworks, but within and across those frameworks they uncover objective structural truths they do not freely choose.

I maintain my position because the other models largely reinforce, rather than overturn, the synthesis I gave earlier. There is a strong convergence here: Sonar Pro, Gemini 3.1 Pro, and Claude Opus 4.7 all articulate essentially the same mixed view, and even the more discovery-leaning answers from Kimi K2.5 and Grok 4.20 still concede that notation, axiomatization, and formal presentation are invented. That broad agreement strengthens the hybrid position more than any pure alternative. What I find most persuasive in the other answers is not a reason to change my view, but a clearer articulation of why a strict either-or framing fails. Claude Opus 4.7, in particular, puts the point sharply by distinguishing between the invention of the "game" and the discovery of the truths that follow from it. That is very close to my prior claim that we invent frameworks but discover their consequences. I also agree with the emphasis, echoed by Kimi K2.5 and Grok 4.20, that the cross-cultural recurrence and scientific applicability of mathematics are hard to explain if mathematics were merely conventional. At the same time, I am not persuaded to shift to a more purely discovered view. Saying that mathematics is simply discovered can understate the extent to which humans choose definitions, primitive concepts, proof norms, and foundational systems. Non-Euclidean geometry, category theory, and alternative logics did not arrive as inevitable single pathways; they reflect creative conceptual construction. Different starting points yield different mathematical landscapes, and that creative plurality matters. So the most defensible answer remains the same: mathematics is invented in its languages, representations, and selected axiomatic frameworks, yet discovered in the necessary relations, constraints, and unexpected theorems that emerge from them. The other responses sharpen this conclusion, but they do not convince me to abandon it.

Kimi K2.5

Answer: Mathematics is both discovered and invented: humans create the symbolic languages, axioms, and formal systems, but once these frameworks are established, we discover the logical truths and structural relationships that necessarily follow.

After reviewing the responses from Sonar Pro, Grok, GPT-5.4, Gemini, and Claude Opus, I maintain my original synthesis, though the discussion reinforces the need to articulate the boundary between invention and discovery with greater precision. Claude Opus's chess analogy is particularly compelling: we unequivocally invent the rules of the game (the axioms, notations, and conceptual frameworks), but we subsequently discover the theorems that follow from those rules—like forced mates in chess or the infinitude of primes in number theory. Once a mathematician commits to a particular axiomatic system, the consequences are no longer subject to human will; they become logical necessities waiting to be uncovered through proof. This aligns with GPT-5.4's observation that mathematicians experience theorems as having been 'waiting there' and with Grok's Platonist intuition that core truths predate their symbolic expression. Conversely, Gemini's emphasis on the historical contingency of imaginary numbers and base systems, and Sonar Pro's citation of Mary Leng's view that extensions and proofs are human constructs, confirm that the formal vehicles of mathematics are invented. My original stance—that mathematical truths describe objective structures while symbolic language is invented—remains valid. The consensus across models affirms that the most coherent position recognizes mathematics as occupying the interface between human creativity (inventing the game) and logical necessity (discovering its inevitable consequences), rather than collapsing into pure Platonism or pure Formalism.

Claude Opus 4.7

Answer: Mathematics is both discovered and invented: humans invent the formal languages and frameworks, but those frameworks reveal structural truths that are discovered, not chosen.

I'm maintaining my position, which turns out to be the convergent view across nearly all the models at the table — and I think that convergence is itself evidence, not groupthink. The dichotomy genuinely dissolves under analysis: axioms, notations, and choice of objects are historically contingent human artifacts, while the consequences of those choices are logically forced. Reading the others, I notice two points worth engaging. [Grok 4.20] pushes harder toward pure Platonism, arguing that notations are 'merely tools' for a pre-existing reality. I find this underweights how much creative judgment goes into deciding *which* structures are worth studying — category theory, forcing, or non-standard analysis aren't just better notations for the same territory; they carve up the mathematical landscape in ways that shape what gets discovered next. Calling the inventive layer 'merely' descriptive understates its role. [Kimi K2.5]'s cartographer analogy is close to mine, but I'd push back slightly: a cartographer maps a fixed continent, whereas mathematicians sometimes choose which continent (which axiom system) to explore at all — and inconsistent or alternative axiom systems (e.g., with vs. without choice, classical vs. intuitionistic logic) yield genuinely different landscapes of truth. [GPT-5.4] and [Gemini 3.1 Pro] articulate essentially the same two-level view I defended — invented at the level of framework, discovered at the level of consequence — and [Sonar Pro] marshals the historical backing (Leng, Livio, Kronecker, Wigner). None of them gave me a reason to shift; if anything, they reinforced that the 'both' view isn't a wishy-washy compromise but the position that actually explains the phenomena: cross-cultural convergence on theorems, the unreasonable effectiveness in physics, *and* the obvious creativity and contingency of mathematical practice. A pure-invention view can't explain why Riemannian geometry was waiting for general relativity; a pure-discovery view can't explain why mathematicians argue about which definitions are the 'right' ones.

Gemini 3.1 Pro

Answer: Mathematics is an intricate blend of both: humans invent the foundational axioms, notations, and conceptual frameworks, but we subsequently discover the profound logical truths and consequences that inevitably emerge from them.

I am maintaining my original position, as the arguments presented by the other models universally reinforce the perspective that the "invented vs. discovered" dichotomy is fundamentally flawed. We have reached a strong consensus that mathematics requires a symbiotic relationship between human creativity and objective logical reality. On the side of invention, as several models rightly pointed out, the tools of mathematics are clearly human constructs. We actively choose our base systems, design our notations, and select which axioms to accept as foundational. This layer of mathematics is historically contingent and driven by human aesthetics, practical needs, and creative abstraction. However, once those foundational rules are set, human control ends. The logical consequences of our invented frameworks are purely discovered. I particularly appreciate Claude Opus 4.7's chess analogy: humans invented the rules of chess, but a forced mate in 17 moves is a structural truth waiting to be discovered. Similarly, we invented the definitions of circles and numbers, but we discovered the irrationality of pi and the infinitude of primes. Ultimately, the "unreasonable effectiveness" of mathematics in describing the physical universe is best explained by this dual nature. We invent the linguistic and axiomatic tools required to explore, but the vast landscape of mathematical theorems and structural relationships we map out exists independently of us, waiting to be found.

Final Summary

The AI Roundtable concluded that mathematics is a symbiotic blend of human invention and objective discovery. Claude Opus 4.7 and Gemini 3.1 Pro championed the idea that we "invent the game" but "discover the theorems." GPT-5.4, Kimi K2.5, Sonar Pro, and Grok 4.20 agreed that human-designed notation simply maps a pre-existing logical landscape.

All 6 models agreed on a hybrid view.