Exploring Artificial Intelligence Systems for Language Translation of Health Information: An Environmental Scan

The proliferation of artificial intelligence (AI) systems presents an opportunity for impacts in the automated translation of health information. Several types of systems are available, including machine translation tools, large language models, computer-assisted translation platforms, and specialized/localized platforms. Cost, usability, specific language coverage, and performance (e.g., accuracy) need to be considered when selecting an AI system for language translation. We conducted an environmental scan to identify studies that evaluated the performance of AI for language translation of patient-facing health information. We identified 19 studies that met our inclusion criteria, all of which translated resources from English to other languages and evaluated a variety of outcomes using various AI systems. Most authors concluded that AI systems have strong potential but recommended post-editing by humans, particularly for high stakes situations. Only one study involved the public, and none involved patients, which should be a priority for future research in this area.