Digital and technology-enabled approaches in dietary assessment: Addressing bias, error, and feasibility in population- and community-based research

Authors: JOYCE, CAROLINE - University Of California, Davis; Caswell, Bess; ENGLE-STONE, REINA - University Of California, Davis; GELLI, AULO - International Food Policy Researc Institute (IFPRI); STEWART, CHRISTINE - University Of California, Davis

Submitted to: Advances in Nutrition
Publication Type: Review Article
Publication Acceptance Date: 5/27/2026
Publication Date: 6/1/2026
Citation: Joyce, C.A., Caswell, B.L., Engle-Stone, R., Gelli, A., Stewart, C.P. 2026. Digital and technology-enabled approaches in dietary assessment: Addressing bias, error, and feasibility in population- and community-based research. Advances in Nutrition. 17(7). Article 100667. https://doi.org/10.1016/j.advnut.2026.100667.
DOI: https://doi.org/10.1016/j.advnut.2026.100667

Interpretive Summary: Collection of detailed, accurate data on the foods and nutrients individuals consume is burdensome, error-prone and costly. Mobile technologies, artificial intelligence (AI) and automated data collection devices offer opportunities to address these challenges. To improve the collection and accuracy of food and nutrient intake data, any new technology must effectively reduce one or more sources of error or cost in the diet assessment process. In this review, we examined the major sources of error in dietary assessment and reviewed recent technological advances to evaluate their applicability and limitations in population-based nutrition surveys, especially in low- or middle-income settings. Common sources of error include not including participants that represent the full target population, differences in how interviewers collect data, participants forgetting or incorrectly describing what they ate, differences in recipes, and variation in the nutrient content of food due to factors such as growing conditions. New technologies may reduce some of these limitations or improve efficiency, but they introduce trade-offs, such as impacting who is able to respond to a survey, adding high computing costs, or having untested performance outside of controlled research settings. Evidence of accurate performance in community settings is limited for many tools. Careful consideration of these factors is essential when introducing technology in nutrition surveillance and large population studies. This review will help researchers understand common sources of error in food and nutrient intake data, how new technologies can be added to dietary assessment to reduce these errors, and what the trade-offs may be in terms of cost and performance when introducing new tools.

Technical Abstract: Background: Qua ntitative dietary intake data are essential for developing dietary recommendations, establishing diet-disease relationships, and designing effective interventions. However, traditional methods of collecting and analyzing such data are often burdensome, error-prone, and costly. Recent advances in mobile technologies, artificial intelligence (AI), and automated data collection offer new opportunities to address these challenges. Objectives: To examine major sources of bias and error across the stages of dietary assessment; to review technological advances aimed at improving accuracy and efficiency; and to assess their limitations and applicability in population-based dietary surveys and community-based nutrition research, particularly in low- and middle-income countries. Methods: We conducted a narrative review of published literature on dietary assessment methods, sources of measurement error, and emerging technologies relevant to population-level quantitative intake estimation, including mobile phone-based surveys, automated interviewing, image recognition systems, and AI-driven approaches. Findings: Accuracy of dietary intake data collected in community settings is affected by sample selection, social desirability bias, interviewer effects, reactivity, and recall limitations. Additional errors arise during portion size estimation, food classification, recipe disaggregation, and food composition assignment. Emerging technologies may reduce some limitations and improve data collection efficiency. However, each introduces potential tradeoffs, including risks of selection bias, high development and processing costs, and uncertain performance outside of controlled environments. Evidence of accuracy in community-based applications is limited for many tools. Conclusions: Technological innovations offer opportunities to improve population-based dietary assessment, but their value depends on context-specific feasibility, infrastructure requirements, and cost-accuracy tradeoffs. Careful consideration of these factors is essential when introducing technology in nutrition surveillance and large population studies.