Foundations of Mathematics and Fundamental Mathematics: Why the Deepest Research Is Often Underfunded

Getting your Trinity Audio player ready...

Foundations of mathematics investigates the logical, conceptual, and structural basis of mathematics itself. Fundamental mathematics develops general theories that may support many other branches, even when they have no immediate application. Unlike ordinary specialized research, both fields work near the roots of the mathematical knowledge tree—and that position makes their importance unusually difficult to measure in advance.

This creates a funding paradox. Foundational discoveries can eventually transform large parts of science and technology, but conventional grant systems frequently prefer projects with short timelines, familiar terminology, established communities, and predictable deliverables.

AI Internet-Meritocracy offers a possible alternative: evaluate mathematical work continuously according to its demonstrated intellectual contribution, downstream use, verification, and influence rather than relying almost entirely on advance predictions by a small grant committee.

What Are the Foundations of Mathematics?

The foundations of mathematics are the theories used to explain or formalize what mathematical objects, statements, proofs, and constructions are.

The term usually includes fields such as:

• mathematical logic;
• set theory;
• proof theory;
• model theory;
• computability theory;
• type theory;
• categorical foundations;
• the study of formal systems and axioms;
• parts of the philosophy of mathematics closely connected to formal practice.

The Internet Encyclopedia of Philosophy defines foundations of mathematics as formal mathematical theories or branches that claim a foundational role for mathematics as a whole. This is distinct from the philosophy of mathematics, which asks broader metaphysical and epistemological questions about mathematical truth and mathematical objects.

For example, foundational research may ask:

• What axioms are needed to derive ordinary mathematics?
• What does it mean for a mathematical statement to be provable?
• Can a formal system prove its own consistency?
• Are some mathematical questions independent of standard axioms?
• Should mathematical objects be represented as sets, types, categories, or something else?
• How can proofs be represented so that computers can verify them?

The foundations of mathematics therefore do not merely constitute another collection of theorems. They examine the language, rules, objects, and standards of inference through which the rest of mathematics is constructed.

Authoritative introductions include the Internet Encyclopedia of Philosophy’s overview of the foundations of mathematics and the Stanford Encyclopedia of Philosophy’s article on the philosophy of mathematics.

What Is Fundamental Mathematics?

Fundamental mathematics is a broader and less formally standardized term. It usually refers to mathematical research that develops deep, general, or widely reusable structures without being directed primarily toward an immediate practical application.

It may include parts of:

• abstract algebra;
• category theory;
• topology;
• number theory;
• geometry;
• functional analysis;
• foundational logic;
• representation theory;
• general theories that unify several existing branches.

A result can be fundamental without being foundational in the strict logical sense.

For example, a new theory of algebraic structures may not attempt to define numbers, proofs, or mathematical truth. Nevertheless, it may be fundamental if it reorganizes existing knowledge, reveals relationships among several fields, or supplies concepts that later become indispensable.

Category theory illustrates this distinction. It is sometimes proposed as a foundation for mathematics, but much of category theory also functions as fundamental mathematics: it provides a general language for structures and structure-preserving transformations across algebra, topology, geometry, logic, and computer science. The Stanford Encyclopedia of Philosophy’s discussion of category theory describes how categories and functors clarify relationships among different kinds of mathematical structures.

Foundations, Fundamental Mathematics, and the Rest of Mathematics

The boundaries are not absolute, but the following distinction is useful:

Type of research	Main question	Typical output
Foundations of mathematics	What makes mathematical reasoning and construction possible?	Axioms, formal systems, proof frameworks, models and notions of computation
Fundamental mathematics	What highly general structures or principles organize mathematics?	New theories, abstractions, classifications and unifying concepts
Specialized pure mathematics	What is true within a particular established domain?	Theorems about defined classes of objects
Applied mathematics	How can mathematics model or solve an external problem?	Models, algorithms, estimates, simulations and optimization methods
Mathematical engineering	How can known mathematics be implemented effectively?	Software, numerical methods and technical systems

A specialized theorem may be extremely difficult and valuable without changing the conceptual organization of mathematics. Conversely, a foundational idea can initially appear simple because its importance lies not in computational complexity but in changing what can be expressed, connected, or proved.

This is why the number of pages, the length of a calculation, or the immediate visibility of an application cannot reliably measure foundational importance.

Fundamental Does Not Mean Elementary

“Fundamental mathematics” should not be confused with elementary mathematics.

Elementary mathematics usually means arithmetic, basic algebra, elementary geometry, and other subjects taught before advanced university study. Fundamental mathematics can be extremely abstract and technically demanding.

Nor does “fundamental” simply mean “important.” The term normally implies that a theory contributes concepts or structures that other mathematical work can build upon.

A useful test is:

A mathematical theory is fundamental when its concepts can become infrastructure for multiple lines of reasoning rather than merely solving one isolated problem.

Why Foundations and Fundamental Mathematics Matter

Foundational and fundamental research can produce several kinds of value.

They reveal hidden assumptions

Mathematicians routinely work within accepted languages and conventions. Foundational analysis makes those assumptions explicit. This can expose inconsistencies, unnecessary restrictions, independence phenomena, or alternative frameworks.

They create reusable intellectual infrastructure

A general mathematical structure can be used repeatedly across fields. Once developed, it may reduce many apparently different problems to instances of a common theory.

They improve formal verification

Logic, type theory, proof theory, and categorical semantics contribute to proof assistants, programming-language theory, and verified software. Research that once appeared remote from application can become essential when mathematical reasoning is encoded for computers.

They reorganize knowledge

Some advances matter because they give better definitions, better abstractions, or better connections—not because they immediately settle a famous conjecture.

A unifying framework may make hundreds of later theorems easier to state and prove. Yet conventional metrics may divide the resulting credit among later applications while assigning little measurable value to the original framework.

They preserve long-term intellectual optionality

No committee can reliably predict which abstract idea will become important decades later. Supporting diverse foundational work preserves multiple possible directions for future mathematics.

Why Foundational Mathematics Is Often Underfunded

There is no single global funding category called “foundational mathematics,” and funding conditions vary by country. Nevertheless, several structural mechanisms place this kind of research at a disadvantage.

1. Its Applications Are Difficult to Predict

Grant applications commonly require researchers to describe expected outcomes, impact, milestones, and beneficiaries before the research has been completed.

That requirement favors projects whose value can already be explained using accepted applications. Foundational work often has the opposite form: it creates a language or theory whose future applications are not yet known.

The more original the framework, the less reliable an advance impact forecast becomes.

This does not mean that foundational mathematics lacks impact. It means that its impact is frequently indirect, distributed, and delayed.

2. It Produces Public Goods

Mathematical knowledge is largely non-rival: one person’s use of a theorem does not prevent another person from using it. Once openly published, it is also difficult to exclude others from benefiting from it.

These are characteristics of a public good. Private investors may therefore capture only a small fraction of the total social value generated by mathematical research.

The OECD states that public support is critical where private initiative alone is insufficient to sustain scientific knowledge and innovation. Its work on public support for research and development reflects the recognized economic need to finance research whose benefits cannot be fully captured by its original producer.

Yet even public funding systems must choose among competing projects. When decision-makers seek visible economic returns, highly abstract mathematics can remain disadvantaged.

3. Evaluation Depends on Recognizable Communities

A grant panel needs qualified reviewers. This becomes difficult when a proposal:

• introduces unfamiliar concepts;
• crosses several established disciplines;
• challenges standard terminology;
• belongs to a very small research community;
• is submitted by someone outside a prestigious institution;
• cannot be classified under a conventional subject label.

Review systems work most smoothly when a proposal resembles research that the panel already understands. Truly new frameworks are difficult to evaluate precisely because there may be no established group of specialists in them.

This creates a circular problem:

1. A field needs funding to develop a community.
2. Funders want an established community before providing funding.
3. The field remains small because it receives little funding.
4. Its small size is then treated as evidence of limited importance.

4. Short Grant Cycles Conflict With Long Mathematical Timelines

A deep mathematical theory may require years of definition-building, error correction, examples, proofs, exposition, and integration with existing literature.

During this period, the work may not generate a steady sequence of fashionable papers. It may instead produce a long monograph, a formal library, or a tightly connected body of results.

Funding and hiring systems often reward frequent countable outputs. They can therefore penalize research whose natural unit is a large theory rather than a series of minimally separated papers.

5. Metrics Favor Visible Downstream Results

Citation counts can detect some forms of academic attention, but they are poor at measuring many foundational contributions.

A foundational idea may be:

• absorbed into standard terminology;
• used without citation;
• implemented in software;
• transmitted through textbooks;
• incorporated into later theories;
• rediscovered under different names;
• important to a small but strategically central field.

The original contribution may therefore function as infrastructure while receiving fewer citations than a popular downstream application.

6. Prestige Becomes a Substitute for Understanding

Because reviewing highly abstract work is difficult and time-consuming, evaluators may rely on proxies:

• institutional affiliation;
• previous grants;
• journal reputation;
• recommendation networks;
• academic rank;
• familiarity of the research program.

These signals are not meaningless, but they can become self-reinforcing. Researchers receive recognition because they already possess recognition, while unconventional or independent mathematicians must first obtain institutional acceptance before their work is seriously evaluated.

This is particularly damaging in foundational research, where originality often appears initially as deviation from the accepted conceptual map.

7. Mathematics Competes With More Visible Research Costs

Mathematical research is sometimes assumed to be inexpensive because it does not always require laboratories or physical equipment.

But mathematicians still require:

• time free from unrelated employment;
• stable living conditions;
• access to literature and computing infrastructure;
• collaborators and specialist feedback;
• editing and formalization;
• conference participation;
• long-term continuity.

The absence of laboratory expenses does not imply that mathematical labor is free.

Public agencies do support major mathematical programs. For example, the US National Science Foundation supports theoretical and applied mathematics through its Division of Mathematical Sciences. Nevertheless, the existence of such programs does not eliminate the allocation problem: only a limited share of proposals can be funded, and unconventional work remains difficult to assess through standard competitions.

The Root Problem: Funding Is Based on Predictions

Most traditional research grants are prospective. A small number of evaluators must predict:

• whether the proposed theory will work;
• whether the researcher can complete it;
• whether the field will find it useful;
• whether applications will emerge;
• whether the project deserves priority over competing proposals.

For ordinary incremental research, such forecasts may be reasonably informed. For foundational research, they are intrinsically unreliable.

A proposal that can confidently describe all its future consequences may not be especially foundational. A genuinely new conceptual system changes the available research landscape, so its most important consequences may be impossible to list before that landscape exists.

This suggests a different funding principle:

Do not ask only whether a committee predicted that a contribution would matter. Measure whether the contribution actually became useful after it was created.

AI Internet-Meritocracy as a Possible Solution

AI Internet-Meritocracy is a proposed system for allocating money to scientists and free and open-source software developers according to measurable contribution rather than institutional gatekeeping alone.

Its relevance to foundational mathematics lies in the transition from one-time prospective selection to continuous retrospective evaluation.

Instead of requiring a researcher to persuade a committee that a new theory will become important, AIIM could evaluate evidence that accumulates after publication.

Possible evidence includes:

• mathematical citations;
• verified dependencies between results;
• use of definitions or constructions in later work;
• independent proof checking;
• adoption in formal proof libraries;
• implementation in software;
• expert criticism and responses;
• successful applications across multiple fields;
• educational use;
• replication or independent reconstruction;
• influence on subsequent theories.

The purpose would not be to reduce mathematical quality to a single citation score. It would be to build a multidimensional and revisable model of contribution.

How AIIM Could Help Foundational Research

Retrospective funding reduces speculative forecasting

A theory could receive funding when evidence of value appears, even if no committee predicted that value in advance.

This is especially appropriate for fundamental mathematics, where utility may emerge gradually and unexpectedly.

Long-tail contributions could remain eligible

A conventional grant decision is usually final. AIIM could continue evaluating a work for years.

A paper overlooked at publication could receive support later if researchers begin using it, verifying it, or connecting it to other fields.

Dependency tracking could reward upstream work

Scientific credit often accumulates at the visible end of a chain. AIIM could attempt to identify upstream dependencies.

Suppose an applied algorithm depends on a theorem, which depends on a general algebraic framework, which depends on an earlier foundational definition. A dependency-aware funding system could distribute part of the value backward through that chain.

This would better reflect how mathematical infrastructure actually works.

Independent researchers could be evaluated by output

Under an output-centered system, institutional affiliation would become one signal among many rather than a prerequisite for visibility.

An independent mathematician could be assessed through publicly inspectable work, proofs, responses to criticism, formalizations, and demonstrated use.

This would not imply automatic acceptance. Unconventional claims would still require rigorous verification. The crucial difference is that exclusion from an institutional network would not end the evaluation before it began.

Negative and corrective work could be recognized

Foundational progress includes more than proving positive results. It also includes:

• discovering contradictions;
• identifying hidden assumptions;
• proving limitations;
• correcting definitions;
• showing that a proposed approach fails;
• establishing independence or impossibility results.

These contributions may be difficult to market as grant promises, but they can save other researchers substantial effort. A continuous evaluation system could reward them according to their actual informational value.

AI Is Not Automatically Impartial or Correct

AIIM should not be presented as an infallible mathematical judge.

AI systems can reproduce biases in their training data, misunderstand novel notation, overweight easily measured activity, and be manipulated by coordinated citation or publication strategies. Mathematical importance also cannot always be inferred mechanically from present usage.

A credible system would therefore require:

• transparent evaluation criteria;
• public audit trails;
• versioned models;
• adversarial testing;
• human challenges and appeals;
• mathematical verification where feasible;
• separation between popularity and validity;
• safeguards against citation rings and synthetic publications;
• plural evaluation models rather than one opaque score;
• governance capable of correcting systematic errors.

AI should assist with large-scale evidence processing, dependency analysis, anomaly detection, and comparison. It should not be treated as an oracle that can independently determine mathematical truth or long-term importance.

The case for AIIM is therefore not that AI is inherently wiser than mathematicians. It is that software can continuously process a much broader body of public evidence than a temporary grant panel can examine.

Funding Work, Not Merely Proposals

The deepest reform proposed by AIIM is a change in what the funding system rewards.

Traditional grants often reward the ability to produce a credible proposal under existing academic conventions. AIIM aims to reward the contribution itself—subject to verification, continuing review, and evidence of value.

For foundations and fundamental mathematics, this distinction is decisive.

Researchers working near the roots of mathematics may not be able to promise an application, identify a large existing audience, or divide a theory into predictable annual deliverables. But once their work exists, it can be examined:

• Is it coherent?
• Are its proofs correct?
• Does it simplify previous theory?
• Does it connect previously separate fields?
• Do other results depend on it?
• Can it be formalized?
• Does it generate productive new questions?
• Does it prevent duplication or error?

These questions are more informative than asking whether the researcher successfully predicted the future before beginning the work.

A Hybrid Model Is More Realistic Than Full Automation

AIIM need not replace every existing institution. A practical system could combine several mechanisms:

1. Prospective microgrants for researchers who need time to develop an idea.
2. Retrospective rewards for completed and verified contributions.
3. Dependency-based payments when later work relies on earlier results.
4. Expert review markets that compensate specialists for careful evaluation.
5. Formal-verification bonuses for converting proofs into machine-checkable form.
6. Public challenges and appeals when an automated assessment is disputed.
7. Long-term re-evaluation as the significance of a theory becomes clearer.

Existing agencies could use such a system as an additional evidence layer. Governments could also explore AIIM as part of a more transparent research-funding architecture, as discussed in the proposal for government science funding through AIIM.

Why This Matters Beyond Mathematics

Foundational mathematics resembles other forms of intellectual infrastructure:

• basic scientific theories;
• research software;
• data standards;
• formal specifications;
• replication work;
• negative experimental results;
• taxonomies and reference databases.

These goods are often essential precisely because everyone can build on them. Yet that same openness makes them hard to monetize and easy for funding systems to overlook.

A society that funds only immediately marketable outputs consumes its intellectual infrastructure without adequately renewing it.

Conclusion

Foundations of mathematics examine the formal and conceptual basis of mathematical reasoning. Fundamental mathematics develops general structures and theories that can support many branches, even when their applications are distant or unknown. The rest of mathematics frequently builds on this infrastructure, but conventional funding systems are poorly designed to recognize it.

The problem is not simply that evaluators dislike abstraction. It is that grant systems demand advance predictions, short-term milestones, familiar classifications, and easily countable outputs. These requirements conflict with the way foundational advances often develop.

AI Internet-Meritocracy proposes a different model: evaluate completed work continuously, trace its downstream dependencies, incorporate verification and criticism, and allocate funding according to accumulating evidence of contribution.

Such a system would not eliminate disagreement about mathematical importance. No responsible system can do that. It could, however, make research funding less dependent on whether a small committee understands a new theory at the exact moment when it first appears.

For foundational mathematics, that change could be the difference between an idea disappearing unnoticed and becoming part of the infrastructure of future knowledge.