Writing a Review Like a Running Log? First Make Cards for Twenty Papers
My first draft of the literature review got six words from my advisor: “This is a list of papers.”
My approach then was: one paragraph on what A studied, one on what B studied, one on C. Each paragraph was correct, but together they had no argument.
Later I changed to making cards first, then connecting the threads.
Step One: One Card per Paper
The card has only four columns, written in a table:
| ID | What it studies | How it was done | What it leaves unresolved |
| 1 | Performance of a material at low temperature | Three comparative experiments | Did not consider humidity |
| 2 | Failure mechanisms of similar materials | Fracture analysis | Small sample size |
| 3 | Effect of process parameters on performance | Orthogonal experiment | No long-term follow-up |
The most important of the four columns is the fourth.The driving force of a review comes from “what remains unresolved,” not “what was studied.”
I often can fill in this column only after reading the discussion, so I usually complete the card after reading the whole paper.
Step Two: Cluster by “What Remains Unresolved,” Not by Year
With twenty cards spread out, I grouped them by the fourth column: those that did not consider environmental factors, those with insufficient sample size, those without long-term follow-up, and those with inconsistent methods.
After grouping, usually only three or four groups remain.These three or four groups are the subsections of your review.
If you end up with more than ten groups, the literature is too scattered, and the topic should be narrowed.
Step Three: Write One Sentence for Each Group About “the Common Weakness of This Group”
For example: “This body of research was generally conducted under laboratory conditions and lacks field data.”
This sentence is the topic sentence of the subsection, and all the papers that follow serve it. If you can't write this sentence, this group should not become a separate subsection.
Step Four: Write “Therefore” Between Groups
The logic of a review is not within groups but between them. At the end of each group, I always write a sentence: “Therefore, ×× remains unclear,” and the next group starts from that question.
When the four groups are connected, they form a line:known → weakness → therefore → next group → another weakness → therefore → my question.
Which Step to Use Tools In
My process is: first search, then make cards, then connect the threads.
For searching and initial screening, I use MuchSci's paper search assistant, filtering by keywords and year to first define the scope of the twenty papers; I fill in the cards myself because “what remains unresolved” requires judgment; I also write the connections myself because that is my argument.
There is only one step where I use MuchSci's chat + agent: I paste the text of the twenty cards into it and ask it to do an initial grouping by “what remains unresolved.” Its grouping isn't necessarily accurate, but itcan point out similarities I missed—once it put two papers I thought were in different directions into the same group, and when I looked back, they were indeed both avoiding the same problem.
Before and After the Revision
· Before: three paragraphs, each listing papers; my advisor criticized it as “illogical.”
· After: four sections, each with one weakness and a “therefore” sentence at the end.
The word count is about the same, but it reads as a line, not a pile of points.
This is also one of my criteria for judging anAI assistant for college students: can it help me turn “a pile of papers” into “a line”? MuchSci is useful at the grouping step, but in the end I have to draw that line myself—because where the line goes depends on what I want to prove.