Reviewing the literature
Literature searches are used to support research, and different types of literature reviews can be pieces of research in their own right. Literature reviews help identify research gaps, determine the scope of a discipline and can synthesise evidence to answer research questions.
GenAI tools can be used to support literature searches by providing overviews of topics and starting references, brainstorming search terms for database searches and mapping connections between authors and between papers. They can provide summaries of topics. Databases may also have GenAI search features which enable natural language searching and provide summaries.
Screening stages of literature reviews can be accelerated by using screening tools which use a semi-automated approach with machine learning. Researchers also evaluate research through the critical appraisal of studies.
Responsible use
- Do read up on the conducting and reporting guidelines for the review that you are planning.
- Consider using AI mapping tools and GenAI to find background information and seeding articles as part of your preliminary search.
- Do appraise GenAI output for potential biases and inaccuracies.
- Do continue to use traditional bibliographic databases for gathering studies for comprehensive literature reviews.
- Consider using semi-automated screening tools to save time.
- Consider seeking suggestions from GenAI for critical appraisal tools.
- Consider using GenAI as a starting point to critically appraise other AI tools.
- Do report how you have used tools in your review for transparency and acknowledge it in your research.
Cautions
- Don’t rely solely on GenAI output. Cross-check against independent resources. GenAI can generate false information and create non-existing references.
- Don’t upload articles or data into summarising GenAI tools. You may be breaching copyright, licences or privacy legislation.
Case examples
Click on the plus icons below to explore exemplar and cautionary hypothetical scenarios for reviewing the literature.
Exemplar - Case example 1
Rashida is writing a short literature review for a grant application to support her research into the social impacts of COVID-19 on primary school students. Her grant application is due next week, and she is feeling under pressure to complete it. She uses GenAI to help her identify key studies already exigent in her field, and then uses the suggested references and background summary as a jumping off point for her own reading. She uses her own critical reading skills in reading and interpreting the research.
Rashida is careful to read the original studies suggested by the GenAI tool as she knows that GenAI tools sometimes hallucinate information, including references. She also confirms all interpretations that GenAI offered and writes her literature review entirely in her own words. Rashida is careful to save her prompts and the output from the tool so she can accurately disclose her use of GenAI if required in any future publications. Rashida has followed the Deakin’s guidelines to support the use of Generative AI for researchers.
Exemplar - Case example 2
Lena is conducting a scoping review. She has a clear question and has conducted her systematic search across 3 different databases. She is now ready to screen the 10,000 records she has found. She and her co-reviewer decide to use Deakin’s semi-automated Living Knowledge Review System to screen the records. She is careful to record her use of the tool to enable acknowledgement and transparency during the write up stage.
Lena is compliant with Deakin’s guidelines to support the use of Generative AI for researchers because she has recorded her use of the tool.
Cautionary - Case example 3
Taylor plans to write a systematic review for one of their papers for their PhD. They have heard that conducting a review is time consuming, so they are looking to use any tools that can streamline the process for them. They fail to follow the systematic review guidelines, and they develop prompts to direct a GenAI to write a systematic review for their research question. The tool provides output, but the tool does not have access to important databases so the output is based on flawed and incomplete data.
Taylor is in breach of Deakin’s guidelines to support the use of Generative AI for researchers because they are not compliant with the systematic review guidelines and have produced an unreliable review, and they are not compliant with author guidelines for the review journal which ban the submission of AI-generated manuscripts.