Opinion: Reimagining the Paper

But papers have since become more complex, negating many of these benefits—and even creating new challenges. Shifting the focus of a paper from narration to empiricism would lead to more robust communications and higher data quality.

Present

Today’s papers are often products of several years of work carried out by teams of scientists. Instead of detailing an observation or experiment, they are collections of several empirical observations interwoven with deductive or inductive logic. Examination of the number of data figures per paper in top biomedical journals revealed a two- to four-fold increase over the past three decades. Today’s papers seem to be gradually morphing into the books they were intended to replace.

The complexity of scientific papers, especially in biological sciences, is possibly a consequence of the rapid increase in scientific knowledge, which has made novel hypotheses increasingly complex. However, it is also likely that this complexity is due to a culture of competition. Papers are getting longer and longer because a complete “story” is now a requisite for publication in many high-impact journals.

The increasing length of papers has led to two primary problems that hinder scientific progress. First, the time it takes to generate enough data to reach the standards of high-impact journals prevents some scientists from sharing important data in a timely manner. The other problem, which is often overlooked, deals with the structure of scientific knowledge.

Papers are the foundational unit of structure in science. The more complex they become, the more difficult it is to discern the empirical data they contain. It is thus becoming increasingly difficult to identify what experiments have already been performed or whether certain findings have been independently verified.

Consequently, information retrieval is far from precise. Even the most sophisticated scholarly search engine suffers from a relatively low precision and recall measure; many query results are irrelevant while related results are often missed. Therefore, scientists struggle to keep up with relevant findings in peripheral fields and are unable to identify important literature when they move to a new field. This issue is only amplified by the fact that there are more than 25 million scientific articles indexed in PubMed (with more than 1 million published annually since 2011). The simpler a scientific paper is, the easier it is to quickly archive, retrieve, and utilize.

Over the past decade, there have been several attempts to increase the speed and robustness of scientific communications. These approaches have focused on either publishing data that is often excluded from journal articles on websites like figshare, or skipping the peer-review process through preprint servers like arXiv. In both cases, the quality of peer review is often sacrificed for promptness and convenience, thus discouraging large-scale adoption.

Future

The introduction of a new medium that disseminates peer-reviewed communications of single empirical observations could address many of the shortcomings of the current publication model.

Scientists could, for example, publish their findings through journal-associated online portals where they undergo peer review. More elaborate “hypothesis papers” could take the place of the current research paper in which various pieces of data are logically connected to produce a generalizable conclusion. This radical shift from publishing hypothesis-driven narratives to pure empiricism would, in my opinion, provide a durable solution to the problems facing scientific publishing.

The fragmentation of the modern paper into its empirical constituents would increase the speed of scientific exchange, increase collective engagement, and introduce a new level of resolution to scholarly information retrieval. It would encourage scientists to publish more reliable data (since every figure would be evaluated independently). The decoupling of logic and observation would also decrease experimenter bias for positive results that fit an overarching hypotheses. Finally, by decreasing the time and effort needed to publish, scientists might be more inclined to publish negative or confirmatory results.

One potential criticism of this model is that it might dampen the competitive edge in science by fragmenting rewards and increasing transparency. Scientific competition has long been established as an essential element for maintaining skepticism, quality control, and a high rate of discovery. However, the level and extent of competition within the scientific community has reached pathological levels. It has led to the production of irreproducible data, fraud, and the rampant exploitation of trainees.

Fragmenting the scientific paper would reallocate publication-associated rewards, allowing for a more even distribution. The approach I propose would not eliminate competition; rather, this approach would beget competition driven by quality rather than reward.

Innovation and discovery are rarely a consequence of individual genius. They are outcomes of collective intelligence constructed through efficient communication. Decreasing the quantum of scientific publishing and shifting its focus from narration to empiricism would increase the speed of scientific dissemination without sacrificing quality.

Ahmed Alkhateeb is a postdoctoral research fellow at Harvard Medical School and Massachusetts General Hospital.