When Assessment Shapes Science: Evaluating Research in Context and over Time

1. Introduction

Research assessment is intended to evaluate scientific contribution, quality, and impact. Yet assessment is not a passive measurement process. Once publication counts, citation indicators, journal metrics, and grant income become linked to funding, promotion, recognition, and career progression, they also become incentives. In this sense, assessment can influence the behaviour and priorities of the researchers it is designed to evaluate.

This distinction is increasingly recognized in contemporary research-assessment reform. The San Francisco Declaration on Research Assessment (DORA) challenged the use of journal-based metrics as proxies for individual research quality [revisited in (Schmid, 2017)], while the Leiden Manifesto emphasized that quantitative indicators should support rather than replace expert judgment (Hicks et al., 2015). The 2022 Agreement on Reforming Research Assessment and the Coalition for Advancing Research Assessment (CoARA) [revisited in (Dirnagl and Pariente, 2024)] subsequently broadened the agenda toward qualitative judgment, responsible use of quantitative indicators, and recognition of diverse research contributions. Most recently, the 2026 Organisation for Economic Co-operation and Development (OECD) report, Reforming Research Assessment for Better Science, further emphasizes that research assessment plays a central role in shaping research priorities, funding, careers, and research culture, while narrow assessment measures can generate perverse incentives and undesirable behaviours (OECD, 2026). These developments establish an important foundation. However, a further question deserves greater attention: what happens when the system designed to evaluate science also influences the science that is produced?

This Opinion proposes that research assessment should be understood as both a measurement system and an incentive system. It further argues that scientific contribution is contextual and temporally dynamic. The significance of a discovery may not be apparent when it is produced and may depend on subsequent scientific developments. The central proposition is therefore that research assessment should evaluate scientific contribution in context and over time, recognizing that current visibility is not necessarily equivalent to scientific significance.

2. From Measurement to Incentive

The Journal Impact Factor (JIF) was developed from Eugene Garfield's citation-indexing work to facilitate information retrieval and characterize citation patterns among journals, not to evaluate individual researchers or articles (Garfield, 1955, Garfield, 1964). Nevertheless, journal-based indicators became widely incorporated into research evaluation. The JIF describes characteristics of a journal; it does not directly measure the rigor, originality, or importance of every article published within it.

Citation indicators have comparable limitations. The h-index combines publications and citations but does not adequately capture disciplinary differences, career stage, authorship contribution, collaboration, or activities that do not produce conventional publications (Hirsch, 2005).

Responsible-assessment frameworks consequently emphasize that quantitative indicators should be transparent, contextual, and interpreted alongside qualitative evidence. The more consequential issue, however, is what occurs when these indicators become attached to rewards. Publication counts can influence promotion. Citations can influence reputation. Reputation can influence funding. Funding supports further research and publications, which can generate additional recognition. These feedback loops can transform descriptive measures into powerful career incentives. Researchers may therefore face an implicit choice between pursuing the most important scientific question and pursuing the strategy most likely to succeed under prevailing evaluation criteria.

Evidence increasingly supports this broader interpretation of assessment. Studies of publication pressure have identified strategic responses to evaluation incentives, including attention to journal reputation and publication speed (Johann et al., 2024). Changes in publication patterns around the United Kingdom (UK) Research Excellence Framework deadlines similarly suggest that assessment environments can influence publication behavior (Groen-Xu et al., 2023). The OECD has now explicitly recognized the potential for research-assessment systems to influence research priorities and behaviors (OECD, 2026). The implication is important: assessment criteria do not merely describe scientific activity; they can help determine which activities become attractive.

3. The Consequences for Scientific Risk

This incentive effect matters particularly for research involving uncertainty. Funding agencies appropriately require investigators to formulate hypotheses, demonstrate feasibility, justify resources, and describe expected outcomes. Limited public resources require accountability. Yet if evaluation disproportionately rewards predictable outcomes and strong preliminary evidence, exploratory research may become comparatively difficult to support.

The discovery of cisplatin illustrates why scientific importance cannot always be specified in advance. In 1965, Rosenberg and colleagues were investigating the effects of electric fields on bacterial growth when they unexpectedly observed inhibition of bacterial cell division in cultures exposed to platinum electrodes (Rosenberg et al., 1965). Subsequent investigation identified platinum-containing compounds responsible for the effect and ultimately contributed to the development of cisplatin as a major anticancer drug [revisited in (Monneret, 2011)].

The lesson is not that discovery should depend on serendipity or that feasibility is unimportant. Rather, some important discoveries originate from observations whose eventual significance could not have been fully predicted at the outset.

Recent evidence from more than 100,000 United States of America (USA), National Institutes of Health (NIH) R01 grants found that grants characterized by scientific risk were renewed at lower rates, particularly highly novel research and work from less established investigators (Azoulay and Greenblatt, 2025). These findings do not demonstrate that peer review is inherently conservative, nor do they imply that risky research should automatically be funded. Scientific risk can reflect either genuine uncertainty or inadequate methodology. The more useful distinction is therefore between scientific uncertainty and scientific inadequacy. Peer review remains indispensable because expert judgment can assess rigor, originality, feasibility, and significance in ways that quantitative indicators cannot. Yet peer review is itself subject to uncertainty, incentives, and potential biases. The objective of reform should therefore not be to replace peer review with metrics or metrics with peer review, but to design assessment systems in which quantitative evidence and expert interpretation complement one another.

4. Scientific Significance Is Temporal

The strongest reason to reconsider how scientific contribution is evaluated may be that scientific significance does not always emerge immediately. Gregor Mendel's experiments on inheritance, published in 1866 [revisited in (Gayon, 2016)], received limited attention during his lifetime and were largely overlooked until their rediscovery around 1900 by three botanists, Hugo de Vries, Carl Correns, and Erich von Tschermak [revisited in (Simunek et al., 2011)]. Only subsequently was Mendel's work recognized as foundational to modern genetics.

Mendel's example illustrates more than delayed recognition. It shows that scientific significance can depend on developments that occur after the original research. Darwin had published On the Origin of Species in 1859, seven years before Mendel's experiments appeared in print, yet the conceptual connection between Mendelian inheritance and Darwinian evolution emerged only decades later [revisited in (Berry and Browne, 2022)].

Scientific contributions can therefore acquire meaning through the accumulation and convergence of knowledge. This creates a fundamental problem for assessment based primarily on immediate visibility. A contribution's citation count, journal placement, grant success, or institutional recognition at a particular moment provides evidence about its current position in the scientific system, but not necessarily about its eventual significance.

Citation indicators naturally accumulate over time, and citation practices differ among disciplines. A recently published article and a decades-old contribution cannot be interpreted using identical expectations. Similarly, a young investigator cannot reasonably be expected to have accumulated the same citation or funding record as an established researcher. Thus, current visibility should not be equated with scientific significance. This temporal dimension is complementary to, but distinct from, the contextual approach already emphasized by contemporary assessment reform. Context asks whether evidence is interpreted appropriately given the circumstances of the researcher and research. Temporal awareness asks whether the contribution is being evaluated at a point when its significance can reasonably be observed.

5. From Responsible Metrics to Responsible Assessment

Contemporary reform has rightly moved beyond the simple question of whether metrics are good or bad. DORA, the Leiden Manifesto, and CoARA emphasize responsible use of quantitative indicators, qualitative judgment, and recognition of diverse contributions. The OECD's 2026 work reinforces this direction and highlights the need for assessment systems better aligned with the changing nature of science (OECD, 2026). However, simply adding more indicators may not solve the underlying problem.

Teaching, mentorship, collaboration, peer review, editorial service, data sharing, software, protocols, infrastructure, reproducibility, and community resources all contribute to science. Yet turning each into a numerical target could reproduce metric-driven incentives in a more elaborate form. The relevant question is not simply how many students were supervised, datasets released, reviews completed, or collaborations established. It is what contribution these activities made in context.

The UK Research and Innovation Résumé for Research and Innovation (R4RI) illustrates this shift by allowing researchers to describe contributions to knowledge, development of others, the research community, and broader users and society rather than relying exclusively on publication-based curricula vitae (UKRI, 2026).

The next step is to add time explicitly to this framework. A mature assessment system should ask: What was contributed? Under what conditions was it contributed? How should its significance be interpreted at this point in time? These questions do not eliminate quantitative evidence. They determine how that evidence should be interpreted.

6. Assessment as Part of the Research Environment

Assessment interacts with the broader infrastructure through which science is funded, published, communicated, and recognized. Open access has expanded worldwide access to scientific information, but its financing can shift financial barriers from readers to authors through article processing charges. Researchers should not be expected to use limited grant resources to cover increasingly costly article processing charges (APCs) at the expense of research, personnel, and infrastructure. When researchers must secure funding both to conduct research and to disseminate it, financial constraints may influence not only which research can be performed but also how widely it can be communicated.

Similarly, scientific communication increasingly involves international and multilingual environments. English facilitates global dissemination but should not become an implicit proxy for intellectual quality. Recent advances in artificial intelligence (AI)-assisted language tools provide an opportunity to reduce this imbalance. When used responsibly and transparently, AI can help researchers improve clarity, organization, and accessibility without replacing scientific reasoning, interpretation, or authorship. The aim should not be to homogenize scientific expression, but to prevent valuable ideas from being undervalued simply because they are articulated by researchers whose scientific thinking develops in a non-English language.

These issues illustrate a broader principle: research assessment cannot be separated entirely from the conditions under which research is produced and communicated. If funding, publishing, recognition, and career advancement are interconnected, then assessment becomes part of the scientific environment itself.

7. Conclusions

Research assessment has entered an important new phase. The field has moved beyond the recognition that individual metrics are imperfect toward a broader understanding that evaluation should be qualitative, contextual, responsible, and multidimensional. The next challenge is to recognize that assessment is also an active component of the scientific system. When evaluation criteria influence funding, promotion, recognition, and careers, they can shape research behavior and potentially influence which scientific trajectories are pursued. This makes assessment more than a retrospective exercise. It becomes part of the environment in which discovery occurs.

The temporal dimension is equally important. Scientific significance may be immediate, cumulative, unexpected, or delayed. Some contributions become influential only after new technologies, observations, or conceptual frameworks emerge. An assessment system optimized for what is immediately measurable may therefore systematically favor research whose value is already visible over research whose importance remains uncertain. The objective should not be to eliminate metrics, replace peer review, or create another rigid assessment formula. Quantitative information remains valuable. The objective is to place measurement within an architecture of informed interpretation that recognizes context, uncertainty, and time.

The central proposition is therefore simple: scientific assessment should not merely ask how much was produced, how often it was cited, or how much funding was obtained. It should ask what was contributed, under what conditions, and how its significance should be understood at a particular point in time. Research assessment should not identify the most measurable researcher. It should create conditions for fair recognition, rigorous evaluation, and sufficient intellectual freedom to pursue questions whose importance may not yet be obvious. Metrics should provide evidence. Context should provide meaning. Time should provide perspective. Expert judgment should bring them together. Assessment should serve scientific discovery—not define it.