Accelerating systematic literature reviews
Researchers spend months screening papers for systematic reviews, delaying evidence synthesis.
Problem
Title/abstract screening of tens of thousands of records is repetitive and error-prone under deadlines.
Current workflow
Two reviewers screen independently in reference managers, resolve conflicts in meetings.
Consequences
Slow guidelines and duplicated effort across labs.
How AI might help
Active learning can prioritize likely includes and explain relevance features for human confirmation.
Limits & risks
Domain shift across topics; still needs dual human review for high-stakes work.
Systematic exclusion of relevant papers; reproducibility concerns.
Alternatives today
Crowdsourcing and simpler narrative reviews.
Crowdsourcing needs training; narrative reviews sacrifice rigor.
Evidence
- Cochrane automation methods overview
Discusses machine-assisted screening with human oversight.
Comments
No comments yet.