| Sumario: | Background: Mainstream and assistive technologies play a critical role in enhancing the independence and quality of life for people with disabilities. However, selecting appropriate technology for specific patients requires AT practitioners to navigate a vast number of available products. This study investigates the application of large language models (LLMs) to assist AT practitioners in identifying valuable product insights and informing technology recommendations. Methods: Keplo, an AI-driven customer review analysis platform, was utilized to evaluate six product categories: three mainstream products and three assistive technologies. Using Amazon product reviews, Keplo's initial analysis provides key product information including user demographics, usage patterns, strengths, weaknesses, and potential design improvements. This initial report was subsequently input into a custom GPT, developed by Keplo, to further extract valuable data. Results: After applying prompt engineering, the GPT generated design considerations, product comparisons, and tailored suggestions for individuals with certain functional limitations, offering AT practitioners a detailed and comprehensive guide for product recommendations. Conclusion: These findings demonstrate that LLMs can effectively identify and extract product insights from customer reviews, streamlining the process of prescribing products for individuals with disabilities. The results also reflect limitations of LLM analysis and the need for AT Practitioners to critically review customer reviews for applicability to their clients. IMPLICATIONS FOR REHABILITATION: The study demonstrates that Large Language Models (LLMs) can effectively analyze customer reviews and provide valuable product insights for both mainstream and assistive technology. By leveraging the tools employed in this article, clinicians can extract detailed information about the strengths, weaknesses, and usability of a cohort of products. This improves the prescription of effective technology for patients with a wide range of functional limitations. The product insights identified through this approach can also inform future improvements in assistive technology product design, ensuring that devices are more user-friendly and suitable for people with disabilities.
|
|---|