How to read survey paper properly and efficiently?

Question

I want to know how to make the best use of survey paper? When I'm new to a particular field, I try to grab some survey paper. However, these papers just outline briefly each methods without going deep enough. And after reading the whole paper I am left with I understand a bit of everything but not deep enough, and I'm still left with a lot of question marks about why certain method is used, why certain modifications are necessary, and how particular approach works in details.

However, digging deep into each of these questions will definitely lead me to lots of rabbit holes that would never end.

This makes me wonder if I'm using survey paper the right way (i.e. to get into a new field)? If not, what are the most common/true purposes of survey papers that serve the research communities? How can I make the most and best use of these brief survey papers?

Thank you.

Answer 1

From your question, I don't see anything that you are doing wrong. If there is a problem as you describe it, then the problem is more likely with the survey papers you are reading than with you.

these papers just outline briefly each methods without going deep enough

By definition, that is what a "survey" paper is: it gives a broad survey of a topic, not a deep-dive into any specific aspect of the topic. However, a survey paper cites relevant references so that a reader like you knows where to go to dive deeper into any specific surveyed topic.

I'm still left with A LOT of question marks about why certain method is used, why certain modifications are necessaries, and how particular approach works in details

This is entirely a matter of the quality of the survey paper. A high-quality survey paper should not only list what has been done, but it should answer the why questions that you are looking for.

If you find the only survey papers that are published to be unsatisfactory for answering the most important questions, then there is not much to be done other than hope that an expert will write a high-quality survey. Otherwise, you might consider this as an opportunity for research so that later on, when you build sufficient expertise, you might write that survey paper that you wish had been available to you.

An intermediary solution might be to contact an editor of a highly reputable journal in the field and suggest that they invite a senior expert to write an appropriate review (you could propose some experts based on the most highly cited papers in the survey). Editors often appreciate such pointers because they want to publish articles that will get widely read, so an editor could take your unsolicited suggestion very seriously.

Answer 2

In my own experience, survey papers are extremely useful for a bird's eye over the particular field: they allow you to know in which subfields current and previous research is structured, get familiar with the main classical as well as state-of-the-art authors, and any special notation commonly used.

However, they are not the best tool for familiarizing with specific methods, specially if the field includes any form of mathematics. Unless the survey is of extremely high quality, most times there is no structured relationship or narrative when presenting the particular methods, there is abuse of notation everywhere in order to avoid overcoming the particular differences among different authors dealing with subtypes of an original method, there is not enough distinction between core contributions and more marginal ones, etc.

So, unless the survey is of high quality, I recommend to read it without being concerned if you don't understand the specifics of each method, taking note of what interests you within the new field, and then look for a book that explains it in more detail, or go the particular paper cited should it be a specific method.

Answer 3

I agree with another answer that you do not seem to be doing anything wrong.

I would like to add that, depending on the area, there must be already a textbook covering the topic. They are usually lagging a few years behind, but they usually present this few-year-old knowledge very comprehensively, so, depending on the field, this might be your second choice after the survey (maybe even your primary choice).

Even if the field is very fast-moving, e.g. AI, where modern ideas didn't exist 5 years ago¹, looking at a textbook can help you build the necessary foundation to avoid the rabbit holes.

^{1. E.g. in this survey, most of the papers are 1-2 years old}