Ethical Use of Artificial Intelligence (AI) in Writing Scientific Literature Reviews

Introduction

Artificial Intelligence (AI) is rapidly transforming how scientific research is conducted. Today, AI-powered tools can search thousands of scientific publications within seconds, organize complex information, summarize research findings, improve academic writing, and help researchers identify patterns across large bodies of evidence. These capabilities can make research faster and more efficient than ever before.

Using technology ethically is a broader challenge that extends beyond academia—it also impacts our daily interactions and emotional health (see my reflections on how to Digitalize Your Anger)

However, AI is exactly that—a tool. It cannot replace scientific reasoning, critical thinking, professional judgment, ethical responsibility, or scholarly creativity. While AI can assist researchers throughout the literature review process, it cannot independently determine whether evidence is scientifically robust, identify subtle methodological weaknesses, or develop original scientific arguments. These responsibilities remain entirely with the researcher.

A high-quality literature review is much more than collecting articles or summarizing published studies. It is a systematic, analytical, and evidence-based process that evaluates existing knowledge, identifies what is already known, recognizes areas of uncertainty, and explains how new research can contribute to advancing science. Responsible use of AI can strengthen this process, but only when combined with human expertise, intellectual honesty, and rigorous scientific standards.

1. Define Key Concepts and Keywords
Every rigorous literature review begins by clearly defining the central concepts within the research topic. A concept is an idea or phenomenon being studied, such as patient safety, diabetes, mental health, or artificial intelligence. A keyword is a word or phrase used to search academic databases for relevant scientific publications.

Researchers should first identify every important concept contained within their research question. Each concept should then be carefully defined using reliable scientific sources such as peer-reviewed journal articles, academic textbooks, international clinical guidelines, or recognized professional organizations.

This foundational process requires researchers to systematically execute the following steps:

• Identify precise keywords, synonyms, abbreviations, and alternative terminology.
• Utilize controlled vocabularies, such as Medical Subject Headings (MeSH), PubMed / MEDLINE, Zotero, EndNote, Mendeley, where appropriate to narrow search parameters.
• Develop effective database search strategies using Boolean operators (AND, OR, NOT).
• Compare alternative definitions proposed across different research circles.
• Evaluate the unique strengths and limitations of these competing definitions.
• Select the definition that best aligns with the overall research objectives.

For example, the term artificial intelligence may be defined differently in computer science, healthcare, education, and ethics. Understanding these differences allows researchers to establish conceptual clarity and avoid ambiguity throughout their review. A clearly defined conceptual foundation ensures that literature searches are comprehensive, transparent, and scientifically reproducible.

2. Build a Strong Theoretical Framework
Scientific research should not exist in isolation. Instead, it should be guided by established theories or conceptual models that help explain why a phenomenon occurs.

A theoretical framework acts as the intellectual foundation of a research study. It provides a structured way of understanding relationships between concepts and helps researchers interpret their findings within the context of existing scientific knowledge.

To construct a resilient framework, researchers must:

• Identify established theories relevant to the specific research topic.
• Evaluate several possible theoretical perspectives to find the best fit.
• Select one or two frameworks that directly support the core research question.
• Explain the core assumptions underlying each chosen theory.
• Discuss their strengths, weaknesses, and current academic criticisms.
• Demonstrate exactly how the selected framework informs the study design, the interpretation of results, and the expected contribution to knowledge.

For instance, a study exploring technology adoption in healthcare might draw upon behavioral science theories to explain why clinicians choose to accept or reject AI-supported clinical decision-making tools. A carefully selected theoretical framework strengthens the scientific credibility of the entire study.

3. Critically Appraise the Scientific Literature
One of the most common misconceptions among new researchers is that a literature review simply summarizes published articles. In reality, a high-quality literature review critically evaluates the quality, reliability, and relevance of existing evidence.

Researchers should examine approximately five to ten key studies during smaller reviews, although systematic reviews and doctoral research often include substantially larger numbers of publications.

For every single study handled, researchers must evaluate:

• What was the primary purpose of the study?
• Which specific research question was being addressed?
• What explicit study design was used?
• Who participated in the research and how were participants selected?
• What were the precise inclusion and exclusion criteria?
• How was the data collected and managed?
• Which statistical or qualitative analytical methods were utilized?
• What were the principal findings?
• What methodological strengths improved the credibility of the results?
• What weaknesses or potential sources of bias may have influenced the findings?

Researchers should also explicitly employ internationally recognized critical appraisal tools whenever appropriate:

• CASP (Critical Appraisal Skills Programme): Evaluates the quality and trustworthiness of qualitative studies, cohort studies, case-control studies, and systematic reviews.
• PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses): Provides internationally accepted standards for reporting systematic reviews and meta-analyses. 
• STROBE (Strengthening the Reporting of Observational Studies in Epidemiology): Improves the quality and transparency of observational research layouts. 
• CONSORT (Consolidated Standards of Reporting Trials): Establishes explicit reporting standards for randomized controlled trials. 

These toolsets help researchers assess whether published findings are truly reliable and applicable to their own research questions.

4. Compare and Synthesize the Evidence
Perhaps the most intellectually demanding stage of a literature review is evidence synthesis. To synthesize means to combine findings from multiple studies into a coherent understanding of what the evidence collectively demonstrates. This is fundamentally different from summarizing studies one after another.

During synthesis, researchers should explicitly identify:

• Areas where findings consistently agree across independent teams.
• Disagreements, anomalies, or conflicting evidence.
• Methodological differences that may explain those inconsistent results.
• Recurring themes and paradigms across the literature.
• Emerging scientific trends over time.
• Strengths of the overall evidence base.
• Remaining uncertainties requiring immediate further investigation.

The objective is to produce an integrated interpretation that demonstrates how the literature, as a whole, advances understanding of the research topic. A well-synthesized literature review tells the story of the collective evidence rather than the story of individual papers.

5. Identify Research Gaps
Scientific research advances because unanswered questions remain. A literature review should clearly demonstrate where existing knowledge is incomplete and explain how new research can address these deficiencies.

Researchers should look for five specific types of gaps:

• Knowledge Gaps: Occur when evidence is limited, inconsistent, or entirely absent.
• Methodological Gaps: Arise when previous studies have used weak research designs, small sample sizes, inappropriate analytical methods, or inconsistent measurement tools.
• Theoretical Gaps: Exist when important theories have not been adequately tested, compared, or integrated.
• Population Gaps: Refer to groups that have been underrepresented in previous research, such as older adults, children, minority populations, or people living in low- and middle-income countries.
• Contextual or Geographical Gaps: Occur when evidence from particular healthcare systems, cultures, regions, or socioeconomic settings is entirely lacking.

Clearly identifying these gaps provides a strong, undeniable scientific justification for conducting new research.

Preventing Plagiarism and Maintaining Academic Integrity
Academic integrity is one of the cornerstones of scientific research. Plagiarism occurs when another person's words, ideas, data, or intellectual work are presented without appropriate acknowledgment. Plagiarism may be intentional or accidental, but both severely undermine scientific credibility.

To safeguard their integrity, researchers must adhere to these practices:

• Read original publications carefully and thoroughly before attempting to write.
• Fully comprehend the evidence before attempting to paraphrase.
• Express ideas using their own distinct scholarly language.
• Acknowledge every source accurately through an appropriate citation style.
• Synthesize information from multiple studies rather than relying heavily on one publication.
• Verify AI-generated summaries directly against the original research articles.
• Use plagiarism-detection software systematically before submission.

Good academic writing is not about changing a few words. It is about demonstrating genuine understanding and presenting evidence through original scholarly interpretation.

Ethical Use of Artificial Intelligence
Artificial Intelligence should be viewed as an intelligent research assistant—not as an independent researcher.

AI can responsibly support researchers by:

• Searching scientific databases more efficiently.
• Generating keywords and alternative search strategies.
• Summarizing long research articles for rapid appraisal.
• Identifying common themes and trends across massive studies.
• Improving grammar, structural clarity, and academic style.
• Organizing references and bibliographies.
• Assisting with evidence synthesis under strict human supervision.

Conversely, AI should never be used to:

• Invent references, DOIs, or citations (hallucinations).
• Fabricate research findings, statistics, or datasets.
• Generate fictional quotations from real researchers.
• Replace human critical thinking and analysis.
• Produce unsupported scientific conclusions.
• Manipulate evidence to align with a bias.
• Claim authorship of scientific work.

International publishing standards make it clear that AI cannot accept responsibility for scientific content and therefore cannot be recognized as an author. Researchers remain fully accountable for every statement, interpretation, citation, and conclusion included in their work.

Final Thoughts
Artificial Intelligence represents one of the most significant technological developments in modern research. Used ethically, it can improve efficiency, enhance productivity, and support researchers in managing the rapidly expanding scientific literature.

Nevertheless, excellent research depends on qualities that no AI system can replicate: intellectual curiosity, critical thinking, scientific skepticism, ethical judgment, creativity, and independent scholarly interpretation.

Ultimately, a literature review is not judged by how quickly it is produced, but by the depth of its analysis, the quality of its evidence, the strength of its theoretical foundation, the clarity of its synthesis, and the originality of its contribution to scientific knowledge. AI may help researchers work more efficiently, but scientific excellence will always depend on the expertise, integrity, and responsibility of the human researcher.

If you enjoyed this deep dive into digital ethics, subscribe to our newsletter below and drop your thoughts on AI in the comments section!

---