Systematic Reviews in Elicit

Systematic Reviews are currently available on Pro, Team, and Enterprise plans.

With AI and language models, Elicit can help you save up to 80% of the time it takes to run systematic reviews, without compromising on accuracy. The Systematic Reviews workflow guides you step by step through search, title & abstract screening, and full-text data extraction, also providing a research report to summarize the most relevant papers in your review at the end of the process.

Elicit now supports almost all major steps of the systematic review process:

Elicit Systematic Review guides you through each of these steps:

1. Gather all of the papers for your systematic review

2. Define your screening criteria

3. Evaluate the screening decisions

4. Define your extraction criteria

5. Extract data from all of your screened-in papers

Start with your research question. Elicit will suggest ways to clarify the question or explore additional angles.

Elicit will use this research question to suggest screening criteria and data extraction fields.

Consolidate all of your papers into a single Systematic Review. Have Elicit search for papers, upload PDFs, or select papers you've already uploaded from your library. You can add up to 500 papers from Elicit's semantic search and up to 500 uploaded PDFs, for a total of 1000 papers per review.

Next, refine your screening criteria on a small sample of papers. Elicit will automatically generate relevant screening criteria (using your original research question), so you don’t have to start from scratch.

At any point, you can edit the AI suggestions, turn them off it, or manually enter your own screening criteria. Click the "Evaluate screening" button when you're happy with your criteria, and Elicit will run the full set of papers against your screening criteria.

Elicit will sort papers by their likelihood of meeting all of your screening criteria, providing a screening score and showing you which criteria the paper met. If Elicit included a paper that you feel should be excluded (or vice versa), you can click on the paper to see the detailed screening recommendations. Click the Override button in the right sidebar to override Elicit's decision for that paper.

You can also increase or decrease the screening threshold in order to include or exclude more papers. The higher you set the threshold, the more papers will be excluded from your paper set:

The screening step currently uses the paper's abstract. The Data Extraction step will utilize the full text for any paper where it is available.

As with screening, Elicit will generate suggestions for data extraction based on your research question. You can edit, override, or manually enter your own.

You'll first define your extraction columns on a small sub-set of your papers. Once you're happy with your extraction columns, click the "Run extraction" button to extract data from the full set of papers. You'll see a popup confirmation button confirming that you'd like to run the data extraction and how many papers will be deducted from your paper extraction allowance.

Elicit always provides supporting quotes from the paper and an explanation so that you can easily check the AI-generated answers for accuracy.

The final step in a Systematic Review is for Elicit to generate a research report summarizing your review. In a Systematic Review that screened in more than 40 papers, Elicit will use the 40 papers with the highest screening score to generate the report. At this time, 40 papers is the maximum number that can be included in a research report in Elicit. The data extraction step is run on the full paper set.

• If you have institutional access to paper databases through your school or work, Elicit's browser extension can use that access to help you pull in the full text for more papers before the Data Extraction step: Elicit's Browser Extension

• You can export the results of any step into a CSV for further data cleaning or analysis.

• You can go back to any step of the process and change it at any time. This allows you to create "living reviews" that you can easily update as new papers are published.

• You can run any step independently without running the others. If you just want to do screening without doing data extraction (or vice versa), feel free!

• The entire review project is shareable. On Elicit Team and Enterprise, you and your collaborators can edit reviews together in real time! You can also share view-only links to the review: Share your workflows