Machine learning and automation in reviews is now a reality, but do we yet know how to use these technologies?

Article type
Year
Authors
Thomas J1, Graziosi S2, Brunton J3
1EPPI-Centre, UCL. 'Transform' project; CQIMG
2EPPI-Centre UCL Institute of Education
3EPPI-Centre UCL
Abstract
Objectives:
1. To introduce technologies for automating parts of the review process including: study selection; risk of bias assessment; and synthesis.
2. For participants to try these technologies for themselves.
3. To discuss methodological issues concerning their use.

Description: In response to the ever-increasing deluge of data, new technologies have been developed to automate some of the review process. It is now possible to ‘teach’ the machine to distinguish between relevant and irrelevant citations for a review; and to classify studies according to their possible risk of bias. It is even possible to automate some of the process of writing the text of the review itself. However, while technologies have been developing quickly, methods to guide their appropriate use within systematic reviews have advanced more slowly. While this is to be expected, technologies for use in reviews are now in place and reviewers are beginning to use them in their reviews. Moreover, the Cochrane evidence ‘pipeline’ is now live, and we need to consider the methodological implications of using it – and other technologies – in Cochrane Reviews.

This workshop will open with a short introduction to the technologies now available to automate parts of the review process. The main part of the workshop will be hands-on trialling of tools on participants' laptops and discussion in small groups concerning their use. We will conclude with all-group discussion.