New article in Ecological Informatics – Harnessing Artificial Intelligence for efficient systematic reviews: A case study in ecosystem condition indicators
As the popularity and capacity of generative AI continues to grow, our new paper looks at how these models can be used in the systematic review process to gain insight into assessing ecosystem condition.
Effective evidence synthesis is important for the integration of scientific research into decision-making. However, with complex concepts and increasing numbers of scientific articles, systematic literature reviews can entail a substantial workload. In our paper, we used the GPT-3.5 model to screen literature for inclusion in a systematic review on ecosystem condition indicators, and found this method increased speed and efficiency compared to traditional processes. Crucially, the AI responses were also able to maintain the high level of accuracy required for a robust review. We also highlighted the potential of AI to help contextualise the complex concept of ecosystem condition, a topic which can be approached from a variety of theoretical perspectives. Through the process of developing an optimal prompt, we confirmed the need to define the multidimensional nature of ecosystem condition. We also identified the importance of being clear on what ecological topics do not meet the key criteria of comprehensiveness when describing the state of an ecosystem. Our paper outlines the great potential posed by the continued evolution of the accessibility and capacity of AI tools, but also the need for careful validation of new methods.
Read the article here: external page https://doi.org/10.1016/j.ecoinf.2024.102819