Research Rabbit

Research Rabbit explores the citation graph around papers you already have. You seed it with a few references and it surfaces what cites them, what they cite, and what tends to be cited alongside them, as an interactive network you can navigate.

The reason it complements rather than duplicates a literature search is that keyword search and citation traversal fail in different ways. Keyword search finds papers using your vocabulary — which misses the ones solving the same problem under a different name, and misses whole subfields that arrived at your question from another direction. Citation structure does not care what anything is called. A paper heavily co-cited with three you already rely on is relevant regardless of whether its abstract shares a single term with your query.

That is exactly the knowledge graph argument applied to literature: relationships that exist implicitly become traversable, and queries nobody specified in advance become answerable.

What it is good for

• Finding the paper everyone in a field cites that you have not read. Co-citation makes this immediately visible, and it is the fastest way to locate the thing you are missing.
• Entering an unfamiliar area. Seeding with two or three papers and looking at what clusters around them gives a map of the subfield faster than reading reviews.
• Tracking forward. Following citations of a foundational paper shows where an idea actually went, which is not always where its authors expected.

Where it does not help

Citation graphs are biased toward what is already well cited, so a recent paper or an unfashionable line of work is under-represented by construction. Using it as the primary discovery method quietly narrows a review toward the mainstream of a field — the opposite of what a literature search is for.

It also cannot tell you a paper is wrong. Heavily cited and heavily criticised look identical in the graph, and the argument against threshold thinking applies here too: a citation count is a continuous, noisy signal about attention, not a verdict about quality, and treating it as the latter is a mistake the interface makes easy.

I use it to widen a search and then read to narrow it. Reversed, it produces a bibliography that looks thorough and is not.

See also: knowledge graphs, ORKG for a structured-claims approach to the same problem, DOIs, and the open science map.