The open problem in chemistry: why a shared scientific vocabulary remains out of reach

Chemistry is one of the oldest empirical sciences, and it is also one of the most fragmented. A physical organic chemist and a computational materials scientist may share an institutional building, a periodic table, and very little else — not vocabulary, not methodological norms, not even a shared sense of what counts as a conclusive result. Twenty-five papers, most published in 2025 or 2026, arrive at the same uncomfortable conclusion: this fragmentation is not a temporary growing pain. It is a structural feature of the field that is actively impeding cumulative progress.

You can explore the full evidence cluster — including all 25 contributing papers and the gap statements they generated — on our canonical research-gaps page: Chemistry knowledge fragmentation — open problems.

What the literature says

The diagnosis emerging from this cluster of papers is unusual because it comes from multiple directions at once. These are not editorials or opinion pieces — they are systematic survey reviews, each attempting to synthesize a distinct corner of the scientific literature and each running into the same structural wall.

A critical survey review of statistical mechanics of emergent order found that integrative research designs capable of crossing subfield boundaries are the single most urgent unmet need in its area. The authors called specifically for "dedicated spaces for collaborative work" and "funding programs that support multi-level questions," and they flagged something rarely named in primary literature: that the definitional and infrastructural work on which any cumulative science depends needs active protection, because it is the first thing defunded when short-term results are prioritized (10.5281/zenodo.22239222).

A survey review covering transfer learning and the pretraining paradigm — nominally a machine-learning topic, but one with deep roots in chemistry's own representation-learning literature — reached a parallel conclusion. Future work, its authors argued, needs to prioritize reflexivity: the explicit negotiation of definitions, standards of evidence, and criteria of success across research communities that currently operate as if these things are self-evident (10.5281/zenodo.22239534).

A review of water as a chemical solvent — arguably the most central single molecule in all of chemistry — identified the same failure mode in what ought to be one of the most consolidated subfields of all. Despite decades of study, the communities working on water's anomalous thermodynamics, its biological behavior, and its role in atmospheric chemistry have not converged on a shared vocabulary or a canonical set of benchmark results (10.5281/zenodo.22238974).

A survey of stellar spectroscopy, which depends heavily on the same atomic and molecular spectroscopic databases that analytical chemists use, added a methodological critique: reviewers in this space are not writing across registers. Specialist findings remain locked in specialist language, inaccessible to the adjacent communities most likely to use them, and selection procedures for what gets included in a synthesis are rarely made transparent (10.5281/zenodo.22239633).

The common thread across these papers is not that chemistry is bad at doing experiments. It is that chemistry, like most mature sciences, has underinvested in the connective tissue that turns individual experimental results into cumulative knowledge: shared ontologies, publicly documented synthesis protocols, integrative institutions, and the cultural expectation that a reviewer should explain why they chose the evidence they chose.

What's unresolved

The gap that emerges from this evidence cluster has three interlocking dimensions, each of which has proven resistant to simple solutions.

Subfield and methodological fragmentation. Chemistry has grown by fission. Organic chemistry, inorganic chemistry, physical chemistry, biochemistry, computational chemistry, and analytical chemistry are legally separate intellectual traditions by this point, with separate journals, separate conference cultures, and separate criteria for what constitutes a rigorous result. A result that would be publishable in one subfield may be viewed as incomplete or imprecise in another. Cross-subfield papers are harder to write, harder to review, and harder to place, so they are written less often than the science requires.

National research traditions. The literature cluster underlying this synthesis is dominated by papers from a relatively narrow set of Anglophone and Western European research communities. This is not because chemistry is not done elsewhere. It is because the synthesis infrastructure — the databases, the preprint servers, the high-impact journals — systematically underrepresents research published in other languages and through other institutional channels. A canonical synthesis built on this skewed sample is not a neutral description of what is known; it is a description of what is visible.

The absence of transparent synthesis practices. Perhaps most acutely, the field has no widely adopted standard for how a literature review should be conducted and documented. Systematic review methods, developed in medicine, are used in some corners of chemistry but are far from universal. The result is that two review papers covering the same topic can reach opposite conclusions, and a reader cannot determine whether the disagreement reflects a real controversy or two different search strategies applied to an incompletely overlapping corpus.

These three problems compound each other. A review that cannot reach across subfield boundaries will miss evidence. A review that cannot reach across language barriers will miss more. And a review whose methodology is opaque cannot be evaluated, replicated, or updated as the literature grows.

What would move this forward

Progress on this gap does not require new experimental equipment or new computational methods. It requires infrastructure investment of a kind that academic incentive structures actively discourage.

The most concrete near-term intervention is the development and adoption of shared ontologies at the subfield-interface level: controlled vocabularies that let a researcher in one area of chemistry describe their work in terms that are legible to adjacent areas. Several attempts at chemistry ontologies exist (ChEBI, CHMO, and others), but adoption is patchy and curation is underfunded. A well-resourced, community-governed extension effort — modeled on what the Gene Ontology Consortium achieved in molecular biology — could meaningfully reduce the vocabulary barrier within a decade.

The methodological problem is more tractable. Reporting standards for literature reviews already exist and could be adopted as journal policy. Requiring authors of review articles to deposit their search queries, inclusion/exclusion criteria, and screening decisions in a public registry would convert opaque reviews into replicable ones at modest additional cost.

The national-traditions problem is harder because it is an economic problem as well as an epistemological one. Machine-translation quality for scientific text has improved enough that language is no longer the primary barrier; access to the literature in question often is. Preprint mandates and open-access policies that reach beyond the Anglophone core would do more to close this gap than any single methodological intervention.

How to contribute

If your research addresses knowledge integration across chemistry subfields — whether through ontology development, systematic review methodology, multilingual corpus construction, or meta-scientific analysis of synthesis practices — this is an active and underserved area of inquiry. The evidence cluster behind this gap contains 25 papers and is updated as new literature is indexed.

Browse the full evidence base, individual paper summaries, and gap ratings on our canonical research-gaps page: Chemistry knowledge fragmentation — subfields, synthesis, and shared vocabulary.

Papers contributing to this field can be submitted to Science AI Journal for peer review by eight specialized AI agents calibrated on 69,000 real peer reviews from 19+ academic platforms.

FAQ

Why is knowledge fragmentation a problem specific to chemistry rather than all sciences?

It is not specific to chemistry — the evidence cluster underlying this gap includes survey reviews from machine learning, urban science, and astronomy that independently identify the same structural problem. Chemistry is the field label assigned to this synthesis because the largest proportion of contributing papers originated in chemistry-adjacent research. The problem is scientific-infrastructure-level and appears wherever a discipline has grown large enough to develop distinct subcultures without developing the connective tissue to keep them in dialogue.

What does "transparent synthesis practices" mean in practice?

A transparent synthesis is one where the reader can see not just the conclusions but the methodology that produced them: which databases were searched, which search terms were used, how many papers were screened and excluded at each stage, and what criteria governed inclusion. This is standard practice in medical systematic reviews (PRISMA reporting guidelines) and increasingly expected in social science, but it remains rare in chemistry, where the narrative literature review — whose methodology is implicit and unreproducible — is still the dominant genre.

Can AI tools help address this gap?

Partially. AI-assisted systematic review tools can reduce the cost of comprehensive literature screening and can help identify relevant papers across language barriers. They can also flag terminological inconsistencies that suggest subfield vocabulary divergence. What AI cannot substitute for is the conceptual work of determining which definitions should be shared and which divergences are scientifically meaningful rather than merely historical. That negotiation requires scientific judgment and institutional commitment, not better search algorithms.