How To Get New Ideas from Research Papers - Part 1 | Research Tutorials with Dr. Sourish

Research ideas don’t come from reading more papers—they come from reading the right parts of papers with a gap-hunting method. With researchers often able to read only about one paper per week, the backlog becomes overwhelming, and the real risk is losing time and missing opportunities hidden in problem statements and assumptions. The core finding here is a practical workflow for literature surveys that turns each paper into a structured hunt for (1) what problem the authors are really trying to solve, (2) what assumptions quietly limit that problem, and (3) what “points to ponder” can become the seed of a new contribution.

The method starts before results, methods, or conclusions. For each selected paper, the reader is instructed to jump straight to the section where the problem is defined and described in detail—then resist reading the rest. Instead, the reader highlights that problem-definition text and immediately drafts a “canvas” of all the sub-problems or dimensions that would need to be solved to fully address the research problem end-to-end. After that initial list, the reader compares it to what the paper itself identifies: which sub-problems match, which the paper omits, and which the paper includes but the reader missed. The gaps that emerge aren’t automatically “novelty.” Each missing or extra dimension must be judged for importance—ideally through discussion with labmates, teammates, or a supervisor—because some omissions don’t matter, while some “new” items may be irrelevant to the larger goal.

A second gap engine targets assumptions. The workflow asks readers to list every assumption the authors rely on, then test when those assumptions hold and when they fail. If an assumption rarely holds, it becomes a potential fault line for a rebuttal or a new direction. The reader then asks whether removing one or more assumptions changes the problem’s nature—often turning it into a new, more tractable, or more meaningful research question. The transcript emphasizes that assumptions can be fuzzy or crisply formalized; making vague assumptions concrete can reveal entirely new dimensions of the problem.

To manage this work across many papers, the discussion highlights features in Raxter.io (referred to as “racks” in the transcript), including “key insights” to collate research goals from a paper, working notes to store gap lists (e.g., “gaps in research problem formulation”), and discussion threads or critique templates for structured questioning. The platform also supports “compare from your collection,” letting readers link the current section to earlier papers in their own library by argument or assumption similarity—helping connect present gaps to past literature.

Finally, the workflow adds a timeline discipline: when working through a set of papers, read them from oldest to newest. That ordering makes it easier to see how problem formulations and assumptions evolve over time, and how communities “upgrade” or twist research questions—turning literature review into a map of maturation rather than a pile of summaries. The payoff is a repeatable loop: for each paper, extract problem dimensions and assumptions, validate which gaps matter, and then summarize the opportunity gaps that remain after the comparison.