Improve survey strategies (Step 1)

[GiggleLiu]

Current Survey Strategies

The AI picks from these strategies during the survey step (Step 1) based on what is known vs. unknown in each iteration:

#	Strategy	When to use
1	Landscape mapping	First iteration default — broad field overview
2	Adjacent subfield	Deep-dive into a neighboring cluster identified in prior iteration
3	Cross-vocabulary	Abstract away jargon, search other fields for the same structural problem
4	Cross-method	Same problem, different computational or experimental approaches
5	Historical lineage	Who tried before, what failed, what changed since
6	Negative results	Search for papers showing what does not work
7	Benchmarks and datasets	What evaluation infrastructure exists

How to contribute

This list is not exhaustive — there are likely more effective survey strategies we haven't thought of yet. If you know of a strategy that works well in your field, or you've noticed a gap in how the survey step explores the literature, please suggest it here!

Some questions to consider:

• Are there survey strategies specific to your research domain that are missing?
• Are there strategies that would help discover overlooked or underrepresented work?
• Should any existing strategies be refined, split, or merged?
• Are there strategies for discovering non-traditional sources (patents, standards, industry reports, preregistrations)?

Please describe:

1. The strategy name
2. When to use it (what gap does it fill?)
3. An example of how it would work in practice