What do we mean when we use the term artificial intelligence? Artificial intelligence is a term that covers an extremely broad range of concepts. In spite of the recent boom, we have been using artificial intelligence tools for years in the form of Siri (Apple), Alexa (Amazon) and many other features in your smart phones and other smart devices.
Artificial intelligence is a technological system that processes information to accomplish a sophisticated task. AI is a broad term developed in the 1950's with narrower terms nested inside, including Machine Learning (1980's), Deep Learning (2010's), and Generative AI (2020's)
Image from Stryker, C. & Kavlakoglu, E. (2024). What is artificial intelligence (AI)? IBM. https://www.ibm.com/think/topics/artificial-intelligence
Before we can use AI tools, we need to understand what those tools are, how they work, and if they are appropriate for the task for which we want to use them.
For systematic and scoping reviews, the methodology calls for transparent and reproducible methods at every stage of the review. Many AI tools do not meet these requirements. For example, results of a Generative AI query will differ each time it is used, and the way in which is generates results is proprietary and hidden. Therefore, generative AI tools are neither transparent nor reproducible.
Similarly, these reviews are designed to reduce bias at each stage of the review. Artificial intelligence has a known problem of incorporating and even amplifying both the biases of the programmers and the biases of the materials on which it was trained.
Artificial Intelligence & Bias
Video by Mary Broussard at Bertrand Library at Bucknell University
Artificial intelligence tools should never be used in place of human intelligence in an advanced review; however, there are AI tools that can be used to aid human intelligence at various stages of a review. For example, Generative AI could be used to aid in brainstorming to refine your research question or identify search terms, but only in combination with your own brainstorming, index term identification, and exploratory searching. Tools like Research Rabbit can help with citation searching once you have identified your included studies. Screening and data extraction tools can be used to speed up these processes, but should not take the place of two human screeners and extractors.
The important thing to remember is that no tool is perfect, and their work should always be double-checked by a human just the same way that in systematic and scoping reviews, multiple people are required to individually screen sources and extract data to prevent mistakes and individual subjective bias.
As with any tool, you should always approach AI with a critical lens and never assume it will give you a correct answer. If you don't know the right answer in the first place, you won't know if your AI tool has gotten it right or gotten it wrong. The following flow chart can help you think through these decisions.
"Is it safe to use ChatGPT for your task?" by Aleksandr Tiulkanov, CC BY 2.0