Research synthesizer
What it does
Research synthesis is the work of combining findings from more than one study into a single account of what you know. Depending on the tradition it goes by meta-synthesis, cross-project synthesis, thematic synthesis, or secondary analysis of your own back catalogue. The hard part is comparability. Two studies can produce the same sentence out of populations that were never alike, and merging them quietly loses that.
This skill works over the transcripts and past reports you hand it. It reads each study on its own terms first and records the population and method behind it, then looks for what repeats. Where two studies point opposite ways it keeps both and writes the contradiction down as a finding in its own right. Every claim carries the study it came from and how many people sit behind it, so anyone reading can follow a theme back to a sentence a real person said.
What you get back
- Themes that appear in more than one study, each with the studies and participants behind it.
- A contradiction register: pairs of findings that point opposite ways, with the population difference that might explain it.
- A provenance line per claim, naming the study it came from and when that study ran.
- Claims resting on a single participant, listed apart so they don't travel as consensus.
- Questions your existing research cannot answer yet, written as candidates for the next study.
- A markdown synthesis for your repo, or a version you can push into a Great Question insight.
How to use it
Example Prompts:
"Synthesise the four onboarding studies in this folder and tell me where they disagree."
"What do we know about pricing objections across everything in /research/2025?"
"Take these two reports and show me which findings only one of them supports."
What it won't do
It won't tell you which of two contradicting studies was right. Where the populations differ it says so and hands that judgement back to you. It also depends on you supplying method and sample context. Give it twelve transcripts with no study metadata and it will still find what repeats, but it cannot weight anything by how the research was run.
Questions
How is this different from affinity mapping?
Affinity mapping works bottom-up over raw observations inside one body of data. Synthesis starts a level up, over findings that have already been analysed, and its main job is holding several studies against each other. Most teams run one and then the other.
Can it synthesise studies run years apart by people who have since left?
Yes, and that is the common case. It reads whatever context each study carries and flags findings that may have aged out, such as a complaint about a flow you have since rebuilt. Anything it cannot date gets marked undated.
Does it handle surveys and interviews in the same pass?
Yes. The two kinds of evidence stay labelled separately in the output, so a survey number and an interview quote never get blended into one claim. It will also tell you when a survey sample is too small to carry a theme it found in the interviews.
