NUS Releases Key Initiatives and Practical Guide to Support Responsible Use of AI in Food Science
The Food Informatics and Artificial Intelligence (FoodAI) research group, at the Department of Food Science and Technology, National University of Singapore (NUS), together with international partners, has introduced five initiatives and a practical guide aimed at strengthening the responsible adoption of artificial intelligence (AI) in food science. The recommendations respond to growing interest in data-driven methods as well as persistent challenges around AI’s transparency and practical validation.
The research team’s assessment highlights several gaps in current research practices. Among 27 representative AI models predicting food chemical flavors, 81% are closed source, and 63% lack experimental validation, limiting reproducibility and practical assessment. Data resources supporting AI development exhibit similar issues: only 20% of 25 food flavor databases allow unrestricted data access, and among 50 food contamination databases, more than 80% lack appropriate quality control. Such limitations reduce the reliability of AI and its further development.
To address these issues, the team and partner agencies propose five initiatives: (1) integrating food science domain knowledge into model design; (2) improving transparency and reproducibility; (3) establishing benchmark datasets; (4) encouraging laboratory or real-world validation; and (5) developing stronger data standards and infrastructure to support sharing and interoperability.
Complementing these initiatives, the team has released a practical guide outlining how food scientists can apply AI more effectively and responsibly. The guide summarizes methods for building reliable datasets that enable new AI application scenarios, including LLM-assisted literature mining and high-throughput experiments, as well as strategies for developing food-specific algorithms, such as multimodal, physics-informed, and explainable models. It also highlights approaches for achieving stronger applications, including model–experiment iteration and validation under real-world conditions. The team has also proposed a checklist that summarizes key steps for conducting reliable AI-driven food research.
Together, the initiatives and guide aim to support the research community in applying AI more effectively, responsibly, and in ways that align with real-world needs in food science.
References:
(1) Zhang, D. Practical guide for food scientists to build AI: data, algorithms, and applications. Food Chemistry 2026, 499, 147281. DOI: 10.1016/j.foodchem.2025.147281.
(2) Zhang, D.; Liu, M.; Yu, Z.; Xu, H.; Pfister, S.; Menichetti, G.; Kou, X.; Zhu, J.; Fan, D.; Rao, P. Domain knowledge, just evaluation, and robust data standards are required to advance AI in food science. Trends in Food Science & Technology 2025, 164, 105272. DOI: 10.1016/j.tifs.2025.105272.