Ethics of Using AI in Evidence Synthesis: Responsible Artificial Intelligence in Systematic Reviews

Artificial intelligence (AI) is increasingly being integrated into evidence synthesis, particularly in systematic literature reviews where large volumes of scientific evidence must be identified, screened, and organized efficiently. As research output continues to expand, AI-assisted tools are helping review teams manage repetitive tasks such as citation screening, duplicate removal, and data organization. 

However, while these technologies improve efficiency, they also raise important ethical questions about transparency, bias, accountability, and scientific trust.

Why Ethics Matters in AI-Assisted Systematic Reviews

Systematic reviews are designed to provide reliable summaries of existing scientific evidence. Their conclusions often inform clinical guidelines, health policy, reimbursement decisions, and future research priorities. Because of this, every stage of evidence synthesis must remain methodologically defensible.

AI systems can introduce risks such as:

• Incomplete retrieval of relevant studies
• Biased prioritization of evidence
• Inconsistent classifications
• Non-transparent decision pathways

Recent methodological studies show that AI tools perform best when used as support systems rather than autonomous decision-makers. Human verification remains essential throughout the review process.

Core Ethical Principles for Using AI in Evidence Synthesis

Transparency in AI-Assisted Research

Researchers should clearly report where and how AI tools were used during the review. This includes:

• Tool name and version
• Task performed by AI
• Reviewer verification process
• Corrections made by human reviewers

Transparent reporting improves reproducibility and allows readers to understand how conclusions were reached. Emerging reporting frameworks now encourage explicit disclosure of AI-assisted review steps.

Human Oversight Must Remain Central

AI can assist with repetitive tasks, but final decisions should remain under expert human control. Critical review stages that require human judgment include:

• Eligibility assessment
• Interpretation of outcomes
• Risk of bias evaluation
• Synthesis of findings

This principle protects scientific reliability and prevents over-reliance on automated outputs. Leading evidence synthesis organizations continue to recommend that AI supports reviewers rather than replaces them.

Reproducibility and Auditability

A major strength of systematic reviews is that methods can be reproduced. When AI is used, every automated step should remain auditable. Good practice includes:

• Saving prompts or commands
• Live tracking (Living SLR)
• Documenting exclusions
• Preserving screening logs
• Recording reviewer corrections
• Saving user activities

This ensures AI-assisted workflows remain scientifically traceable.

 Data Privacy and Responsible Tool Selection

Some AI systems process uploaded documents or unpublished research materials. Researchers should therefore choose secure platforms that protect:

• Confidential data
• Unpublished evidence
• Institutional research materials

Only validated and institutionally acceptable tools should be used in formal evidence synthesis projects.

Best Practice for Ethical AI Use in Evidence Synthesis

A responsible workflow combines AI efficiency with reviewer expertise. Recommended approach:

• Use validated AI tools only
• Verify all outputs manually
• Report AI use transparently
• Preserve methodological consistency

Maintain human accountability throughout

The Future of AI Ethics in Evidence Synthesis

AI is expected to become increasingly integrated into systematic review workflows, particularly as evidence volumes continue to grow. The future challenge is not whether AI will be used, but how to ensure that its use strengthens scientific quality rather than weakens it. Ethical implementation depends on balancing:

• Efficiency
• Transparency
• Reproducibility
• Fairness
• Expert judgment

Conclusion

The ethical use of AI in evidence synthesis is built on one principle: AI should improve efficiency without replacing scientific responsibility. When used transparently and under expert supervision, AI can strengthen evidence synthesis while preserving the trust that systematic reviews are designed to provide.