The Pulse Report

Is Science Becoming More Productive but Less Curious?

Science has become extraordinarily productive. Researchers can investigate biological processes at the molecular level, generate enormous quantities of data, and use artificial intelligence to analyze information at speeds unimaginable just a few years ago.

By almost every measure of research activity, scientific capabilities are expanding. But what do those measures actually reveal about scientific progress?

We risk confusing the ability to produce more knowledge with the ability to advance scientific understanding.

Producing reliable knowledge is essential to science. But, understanding also advances when a discovery changes how scientists interpret existing evidence, reveals a previously overlooked relationship, or leads researchers to ask a different question.

A discovery in Alzheimer’s research illustrates how new knowledge can change the direction of scientific research.

When Knowledge Changes the Question

In 2012, two independent research teams identified a rare genetic variant associated with a substantially increased risk of Alzheimer’s disease. The gene, TREM2, was already known to play a role in immune-cell function.

At the time, much of Alzheimer’s research focused on amyloid plaques, tau proteins, and the biological processes associated with neurodegeneration. Scientists also investigated inflammation, but the TREM2 findings provided key genetic evidence linking immune function to the disease.

That connection expanded research into the role of microglia, the brain’s immune cells, and how their behavior might influence the development and progression of Alzheimer’s.

The genetic findings gave researchers another reason to examine how these cells respond to damage, remove cellular debris, and potentially contribute to the processes involved in neurodegeneration. The findings strengthened the case for investigating whether the brain’s immune response was playing a role in the disease itself, rather than simply responding to damage already underway.

The discovery did not provide a cure or resolve the many uncertainties concerning the disease. Its importance was partly in how it changed the questions researchers could pursue.

Advances in genetics produced evidence that could be interpreted through existing immunology and further investigated through neuroscience. Specialized research in different areas became relevant to a shared problem.

Deep specialization enables researchers to develop knowledge and methods that would otherwise be impossible. Discoveries in one specialty can also take on new significance when viewed alongside knowledge from another.

This interaction between specialized investigation and wider exploration has long driven scientific progress. A discovery in one field can open questions that require new forms of specialized research in another.

Today, emerging technologies may make it possible to explore relationships across specialized knowledge on a scale not previously possible.

The Paradox of Scientific Productivity

In January 2026, a study published in Nature examined 41.3 million scientific papers to investigate how artificial intelligence is influencing research.

Scientists whose research involved AI published approximately three times as many papers and received nearly five times as many citations as researchers who did not. They also tended to become research project leaders earlier.

By conventional measures of scientific productivity and professional success, the advantages were substantial.

Yet the study found something else. AI adoption was associated with a 4.63 percent contraction in the collective range of scientific topics being investigated and a 22 percent decline in engagement between scientists.

The productivity findings concerned individual researchers, while the narrowing of research topics was seen across the scientific literature. They measure different things, but together they raise an important question about the direction of scientific research.

The researchers observed that AI-assisted work tended to concentrate in areas where substantial datasets already existed, making established problems easier to investigate.

The findings do not establish that AI narrows science. They do, however, suggest that technologies that expand scientific study may be especially useful for accelerating research in familiar areas.

Established research problems often have an advantage. They come with accumulated data, tested methods, and clearer ways to evaluate results. AI can make that work substantially more productive, accelerating investigations that might otherwise take years.

But what happens to questions without well-developed datasets, accepted methods, or a clear disciplinary home?

An unexpected relationship between immunology and neuroscience may be difficult to investigate at first. The available data may have been collected for different purposes. Researchers may disagree about which methods are appropriate or even how to frame the question.

Such work carries uncertainty, and uncertainty is difficult to measure in advance.

If emerging technologies make established lines of research more productive, scientists may have even stronger reasons to pursue them. Meanwhile, some of the questions that could open new areas of understanding may remain difficult to investigate.

How much of this tendency reflects the technology itself, and how much reflects the research system in which scientists are using it?

The Incentives Behind the Questions

How scientists use technology is influenced by the institutions in which they work and the incentives that shape their careers.

Scientists build careers through publications, citations, grants, and peer recognition. Universities organize expertise through departments and disciplines. Funding programs and journals set expectations for what counts as a valuable contribution.

These systems have supported extraordinary achievements. Scientific rigor depends on knowledgeable communities that can evaluate methods, challenge findings, and build on reliable evidence.

They also shape researchers’ choices about where to invest their time and which uncertainties are worth pursuing.

Consider a young scientist who recognizes a possible connection between two fields. Investigating it may require learning unfamiliar methods, developing relationships outside an established professional community, and spending considerable time exploring a hypothesis that may prove wrong.

A more conventional project within an existing specialty may present a clearer route to publication, funding, and professional advancement.

For decades, researchers and scientific organizations have questioned whether promotion, tenure, and funding practices adequately support cross-disciplinary work. Many institutions have responded by creating interdisciplinary centers and joint research programs.

If AI makes it easier to generate publishable results within established research areas, it may strengthen incentives that already favor those areas. Individual scientists can become more productive and successful even if the collective range of scientific research does not expand.

Consider how these incentives may reinforce one another. A researcher applies AI to an established problem where reliable data are readily available. The work produces results that can be published, cited, and used to support additional funding. That success makes the research direction attractive to pursue further.

Much of science advances through the cumulative investigation of established questions. But if the same incentives consistently make unfamiliar questions harder to pursue, scientific institutions may overlook opportunities that don’t fit existing patterns of research success.

Publications and citations provide important measures of scientific activity and influence. But they capture less of the value of an unexpected question, an initially unsuccessful experiment, or a connection that eventually changes how an entire problem is understood.

The TREM2 findings illustrate that distinction. The genetic association was an important scientific result, but its broader significance included the questions it helped open about immune function and Alzheimer’s disease.

The papers that followed can be counted. It is harder to measure the value of changing what scientists believe is worth investigating.

Developing the Ability to See Connections

No individual scientist can master the extraordinary range of knowledge now being produced. Scientific education may need to give greater attention to how researchers learn to recognize when knowledge from another field changes the questions worth asking.

AI can help researchers examine unfamiliar literature, compare findings across disciplines, and generate hypotheses that might not emerge from working within a single field. New computational and experimental tools can help investigate those possibilities.

Access to more knowledge does not necessarily mean knowing what to make of it.

That requires curiosity, experience, judgment, and a willingness to test ideas against evidence. An algorithm may identify a pattern worth investigating, but identifying a pattern is not the same as explaining it. Researchers still need to understand the underlying science and determine whether a proposed relationship survives rigorous examination.

Scientific education understandably concentrates on building disciplinary expertise. Alongside that depth, researchers could have more opportunities to examine unfamiliar evidence, question assumptions outside their specialties, and test hypotheses that draw on several areas of expertise.

Institutions could encourage this by giving researchers room for exploratory work whose value may not be apparent through conventional measures of productivity.

A scientist who recognizes a promising connection does not need to become an expert in every field involved. But that scientist does need enough understanding to see why the connection matters, enough curiosity to pursue it, and access to others who can help determine whether it holds up.

As technology expands access to scientific knowledge, recognizing and testing important connections may become as important as retrieving information.

What Will We Recognize as Progress?

The technologies now entering scientific research offer extraordinary opportunities. They can accelerate experiments, improve analysis, and help scientists investigate questions that were previously beyond their reach. They may also allow researchers to recognize relationships across bodies of knowledge that have developed largely within separate specialties.

But the questions scientists choose to ask, and the research their institutions encourage them to pursue, will shape the possibilities these technologies create.

A research system could become increasingly successful by its established measures, producing more publications, attracting more citations, and conducting experiments faster and with greater precision. Those achievements would be real and valuable. Yet they would not necessarily reveal whether scientists were recognizing relationships that could change their understanding of a problem or opening promising new directions of inquiry.

Scientific tools are expanding what researchers can investigate, while the measures used to evaluate their work continue to emphasize what they produce.

The question is whether the ability to recognize important connections, challenge established explanations, and identify new directions for investigation is advancing alongside the capacity to generate scientific knowledge.

The risk is not that science will produce too much knowledge. It is that measures of scientific success may become increasingly impressive without revealing whether researchers are recognizing the questions that could change what they understand.

Deborah Westphal is the author of The Pulse of Tomorrow (opens in a new tab) and CEO of the Karen Toffler Charitable Trust. More about Deborah →

Continue reading

All Pulse Reports →