The Scientific Paper Needs To Change
When it comes to paper writing, there is a simple recipe. Academic publications typically follow this established structure:
- Title + Abstract
- Introduction
- Related work
- Method
- Results
- Discussion
- Conclusions
- References
This structure is archaic and outdated. In the following, I look at the value-added of sections 1–7.
Title + Abstract + Introduction
These are the most important parts of the paper.
The abstract briefly summarizes the paper and highlights its importance.
The introduction explains the motivation for the work, explains what was done, and summarizes the findings.
A modern paper will still need this section.
Related work
Researchers are facing a data deluge — it is no longer possible to have a complete overview of everything that is written in a given field (unless the field is very small). The related work section thus will necessarily be incomplete and always only “good enough” to satisfy reviewers and pass peer review.
That means we might as well get rid of this section. And let the peer reviewers, in their function as gatekeepers of a certain domain [1], figure out if the publication is original.
The modern publication does not need a related work section. Some related work can be mentioned in the introduction — many researchers already do this, because it provides a nice motivation for the work. Other than that, the related work section can go.
Method
This is an important part of the paper that helps the reader understand what was done. Some patterns, standards, and best practices have emerged on how to write this section. Which means the section is a good candidate for automation. Given a few pieces of key information (type of study or experiment conducted, number of participants, etc.), a machine learning tool could surely do a formidable job in writing this section.
Results
If said machine learning tool can write the method section, then it probably could write the results section as well. Just hand the tool the raw data, the above key information for the method section, and the tool could even analyze the data based on the given instructions.
Discussion
This section allows the writer to conjure up one’s imagination. Design recommendation, implications for the future of X, you name it…
Conclusions
Many readers jump from abstract to introduction and then conclusion to assess if a paper is relevant and important. But the conclusion does not add much value here, because the same information can be found in the abstract and the introduction — it is just regurgitated in this section. Complete waste of space and time.
The publication of the future
Essentially, writing scientific papers is obsolete. The modern paper writer should be supported by machine learning tools, and the writer should only have to include the following sections in the paper:
- Abstract
- Introduction
- Method
- Results
- Discussion
The conclusions can go, as does the related work section.
Much of the writing of these sections should be automated.
Just give me a machine learning tool that I can feed
- the raw data,
- the data processing procedure.
Given this information, the tool could then produce the results. The tool should automatically spew out the method section and write the results section for me. I’d worry about writing the abstract, the introduction, and the discussion.
Any why stop there. Much of the paper should be written in bullet point lists. That would be would be less of a hassle for non-native-speaking researchers, much more inclusive, and more time could be spent on research instead of writing.
[1] Csikszentmihalyi, Mihaly. 2014. The Systems Model of Creativity. Springer.
