At Covidence, we are committed to ensuring that any automation introduced into our workflow meets the highest standards of accuracy, transparency, and ethical considerations. Our approach is built on the following key principles:

Core Principles

Highly Accurate

Automation must maintain or improve the accuracy of systematic reviews conducted on Covidence. We rigorously evaluate automation performance to ensure it meets high-quality benchmarks, allowing teams to trust and rely on Covidence for evidence synthesis.

Where automation features have limitations, these will be clearly communicated in the product and associated knowledge base articles.

Appropriately Evaluated

We develop and assess automation models using high-quality, adequately sized datasets. We strictly separate training, testing, and evaluation datasets to avoid bias and ensure robust model performance. Once our models are operational, we implement continuous evaluation processes to monitor their performance in real-world conditions, allowing us to refine and improve them over time based on ongoing feedback and data.

We’d love for you to share your thoughts and experiences with us! Drop us an email at support@covidence.org with any feedback or observations. Your input helps us make our automation features and models even better.

Human-Centred

We believe users should be able to understand and assess the tool to make informed decisions about its use. That’s why we prioritise transparency, ensuring users can see what AI-driven decisions are made.

Reviewers remain in full control of contributions from automation features, with all automations clearly displayed and easily manually overridden. This keeps human expertise at the heart of decision-making.

Careful Handling of Your Information

We take the protection of your information seriously, ensuring our automation features comply with copyright laws and keep your data secure. Ethical data handling is central to how we work.

To avoid unintended exposure, we carefully manage how data is processed. Our system uses only the necessary information to generate tailored extraction suggestions for your review. Personal details about users are never included.

These suggestions are customised to your content and shared only with your review team - no one else. We rely on trusted large language models that don’t store or share anything from your uploaded files, keeping your work private and reducing the risk of breaches.

Aligned with Industry Standards

We are committed to aligning our AI development with the RAISE guidelines wherever possible, ensuring that Covidence meets the expectations and best practices of the systematic review community. Our proactive approach to compliance keeps us at the forefront of evolving standards and regulations, as we actively stay up to date with industry advancements and contribute to shaping the future of these standards.

Additionally, we work closely with trusted organisations, including Cochrane, to achieve endorsement of our automation features, collaborating to ensure our tools meet the rigorous demands of evidence synthesis and enhance the systematic review process.