Establishing Automated Analysis Systems in the Animal Fluency Task Reflecting Korean-specific Clustering Criteria

Objectives: This study aims to specify the scoring criteria for the animal fluency task (AFT) and establish Korean-adjusted animal clustering criteria. Additionally, we aim to develop an automated analysis approach for the AFT and make this tool accessible for practical use following validation. To validate this approach, we seek to identify the relationship between automated and manual analysis methods and examine the differences in correct responses, switchings, and mean cluster size between younger and older adults. Methods: 275 healthy younger adults and 374 healthy older adults participated in the study. The AFT was conducted within a 1-minute time window, and data were analyzed for the number of correct responses, switchings, and mean cluster size. The study proposed five sub-criteria for scoring correct responses and specified animal clustering criteria into two upper-level and 14 lower-level categories. Additionally, a comprehensive list of animal examples corresponding to each criterion. Manual analysis was conducted by 14 graduate students, and the automated analysis is publicly available on GitHub. Results: The findings revealed significant positive correlations between automated and manual analyses for correct responses, switchings, and mean cluster size. Younger adults also produced more correct responses, exhibited more switching behaviors, and had larger mean cluster sizes than older adults. Conclusion: This study establishes integrated scoring and clustering standards for the objective analysis of Korean participants' performance on the AFT. Furthermore, by providing an automated analysis approach, we contribute to the development of practical resource use in both clinical settings and research areas.