4 main ideas that follow the AI extraction and Evaluation of AI extraction projects.

[jdkent]

1. User experience research
   qualitative replication (what are the steps it takes to complete a meta-analysis)

2. Quantative evalution of AI extracted information
   recall/precision of studies

• how well does semantic search work
• how well does extraction work (evaluation)
• partitioning coordinates to analyses

3. automation using extracted information
   can we create a pipeline using automatically extracted information
   (auto select filters/and selects analyses for you)
   we want this to be efficient (computationally) and low cost

4. Fully autonomous LLM based workflow (similar to otto-SR) for curation of studies.
   use the studies in neurometabench,
   otto-SR in the browser, sub-select analyses that are relevant.