How important is publishing state-of-the-art surveys?

Question

I am a Ph. D. student in computer science. I often cite surveys and find them very useful. By surveys I mean cataloging the state-of-the-art of all scientific efforts made in a single area. Not necessarily drawing conclusions from them nor going farther, though it could be. However my advisor is not keen on working on surveys. This leads me to the following questioning:

1. Is publishing surveys scientifically relevant?
2. Could it be scientifically irrelevant but interesting in my academic development?

Tangent question: is there a different approach to working on and publishing surveys than regular papers? Can I publish surveys in any regular journal or conference?

Answer 1

Often, in CS, survey of state-of-the-art equate review papers with underdeveloped analysis/discussion. Their value to the body of knowledge are eroded.
Reviews are critically important in contributing to the body of knowledge (when they so).

---

It is not uncommon for students (and researchers) in computer science to punt the phrase state-of-the-art. In my view, it gets often abused.

One often sees summary collation of articles/previous works. These sort of 'survey' are nothing but reviews in the true sense of things. At least, that is how I presume they will be classified in most journals.

Is publishing surveys scientifically relevant?

Yes they are. Take for instance in IS (information systems), a closely related sibling), reviews are often used to take stick if the past and forge a direction forward.

Could it be scientifically irrelevant but interesting in my academic development?

In computing, they are relevant. For instance, one gets to know how an algorithm got to be in the first place, how it evolves over time, what others are innovating with it and what possible direction is evolving. At times in computing, algorithms, new models, theories et al are worked on in tangent. Not that that is completely bad. IMHO, synergy and add-on could be beneficial. Obviously, that doesn't stop new approach.

Tangent question: is there a different approach to working on and publishing surveys than regular papers?
Can I publish surveys in any regular journal or conference?

Some journals welcome reviews. So yes, they can be published in journals (as well as accepted in conferences).

One question one should oneself is, what contribution am I making? What am I contributing to the body of knowledge?

The collations (the survey) is more like data/findings that deserves analysis/discussion. Those state-of-the-art are the state of things that should be discussed.

• what is the art to non-discipline person and to CS folks
• what was the state of things and what is the state of things
• what can be postulated about the art and future state
• what is the golden thread
• what is/are the new frontier(s) of the art?

Survey of state-of-the-art shouldn't be left as collation. They should be engaged as reviews as it's done in other fields of research.

---

I'll end with reference to a frontline CS journal on survey of state-of-the-art - the ACM Computing Survey journal

ACM Computing Surveys
2021 Impact Factor: 14.324 (ranked 3/109 in Computer Science Theory & Methods)

• These comprehensive, readable surveys and tutorial papers give guided tours through the literature and explain topics to those who seek to learn the basics of areas outside their specialties in an accessible way. The carefully planned and presented introductions in Computing Surveys (CSUR) are also an excellent way for researchers and professionals to develop perspectives on, and identify trends in complex technologies.
• Contributions which bridge existing and emerging technologies (such as machine learning) with a variety of science and engineering domains in a novel and interesting way are also welcomed.
• Contributions are intended to be accessible to a broad audience, featuring clear exposition, a lively tutorial style, and pointers to the literature for further study

Information and guidelines for reviewers

• Papers for ACM Computing Surveys must be of high quality ...
• There are four main ingredients to an acceptable paper:
  • Technical quality is high.
  • Relevance to significant areas of research or practice is high.
  • The level of general interest is high.
  • The presentation is effective.
• Few papers excel in all of these, but a substandard level in any is sufficient ground for rejection.
• papers devoted primarily to presenting a survey or tutorial that include a new result to organize and convey the material in a better way are within the scope of the journal.

Answer 2

Reviews are very relevant. It is rare that a single paper solves a big problem (it is common to afterwards assign that honor to a single paper, but that is another problem). In practice, many people and groups work on that problem, publish partial solutions, and in doing so collectively move forward. In this process it is very important to regularly take stock. This is what a review is for.

However, this also means that reviews (that get published) are often written by people who were closely involved in this process. They know what happened, not just what got published. They tend to have a better overview of the field than an outsider. So it is unusual for a PhD student to get a review published, which could be an explanation of why your advisor is not keen on you trying this.

In the end, if you want to know what your advisor is thinking, you will just have to ask her/him/them.

Answer 3

I'm a computer science PhD candidate graduating soon that published a review article. What I thought would be a relatively simple process turned into a massive effort, where I spent hundreds of hours writing and shepherding the manuscript through revisions. Writing a review for yourself is one thing; writing a review that others would want to read is totally different. It probably didn't help that I work in an interdisciplinary area, so there were multiple perspectives I had to address.

However my advisor is not keen on working on surveys.

My advisor was ambivalent about me doing the review, and having gone through it I am too. I'm biased, but I think my review article was ultimately worth it and provides a unique contribution to the field. However, it consumed way more time than anticipated, taking my attention away from doing more "traditional" CS research that is typically expected for graduation (e.g., developing methods/algorithms). This has made things slightly more complicated for justifying that I should graduate to some members on my dissertation committee.

is there a different approach to working on and publishing surveys than regular papers? Can I publish surveys in any regular journal or conference?

Yes, surveys/reviews are different from regular papers. Journals will typically have information on the types of papers they accept on their website.