Artificial Intelligence has become an integral part of everyday life, offering virtual assistance for nearly every need, from AI-powered health monitors to educational tutors and nutrition coaches. For instance, Nutrition Coach AI: Diet App was designed to provide users with personalized nutrition coaching, monitor food, track the user’s nutrition habits, recognize food from photos, track calories, analyze medical tests, and offer healthy food recipe recommendations.
AI nutrition coaches use machine learning algorithms, which the developers train using mostly publicly available data on the Internet. AI nutrition coaches can process the data provided by the user, with reference to its training data, to generate insights and actionable, real-time feedback.
The suggestions offered by AI nutrition coaches are always personalized, depending on what the user wants to achieve. For example, the suggestions offered to a user that wants to lose weight will be different from the suggestions it will offer to another who wants to lower their blood sugar level..
While AI nutrition coaches have their benefits, their accuracy in the Black community remains debatable due to bias in training data, which often leads to inaccurate recommendations. AI nutrition coaches often mislabel soul foods and other local African and Black diasporan delicacies. This limitation has been reported in other AI-based health agents.
The Black community, soul food, and AI nutrition coaches
Soul food is one of the delicacies that is linked to the history of Black community. The term was first used in 1964 during the rise of “Black pride” and the celebration of the Black contribution to the American lifestyle. The origin goes back to the rural South when emancipated slaves used whatever they could find to create flavorful delicacies. During the Great Migration, they carried these recipes to the North and all over the United States.
Soul foods mostly include fried ingredients, processed meat, sweetened beverages, and added fats. Several publications have highlighted how these ingredients contribute to the increased risk of heart and kidney diseases, diabetes, stroke, and cancer in the Black community.
It is unlikely that AI nutrition coaches will recommend soul food, considering that the vast majority of publications on the Internet—used in training the AI models—focus on the potentially negative impact of the ingredients used in soul food preparation, notwithstanding that there are publications with suggestions on how to make this delicacy healthier.
Limitations of AI nutrition coaches with native Black dishes
The vast majority of the currently available large language models (LLMs) are trained on data that doesn’t provide cultural context. When these LLMs are used to train AI nutrition coaches, these virtual coaches will automatically inherit these limitations, which can manifest as the mislabeling of nutrient-dense dishes like okra stew and collard greens as unhealthy. Below are some of the factors linked to the limitations of AI nutrition coaches in recommending Black dishes.
1. AI diet apps are built on Western nutrition models
There has been a glaring lack of representation in the development of AI-powered nutrition tools. Registered dietitian, Nonyia Ezinna, identified this gap during his internship in a teaching hospital in Nigeria. The modeling of AI tools on Western dietary patterns excludes the rich diversity of African and Black cuisines. This gap has prompted some researchers to work on designing an African food recommendation system for weight loss.
2. Incomplete training data
The choice of data to train AI models often subtly reflects the racial bias of the developers. Since most developers of AI nutrition coaches are whites, their data is unconsciously dominated by foods that appeal to them. The inclusion of nutrition data from the Black community and other racial minorities usually come as an afterthought.
3. Ignoring cultural context of soul food preparation
Every culture evolves over time. Soul food has evolved with the Black community over several decades. This evolution is rarely captured in most recent data. Instead, the nutritional content soul food is continually judged based on its history.
4. Limited representation from the Black community
The majority of AI nutrition coaches are developed by a small monocultural team. Furthermore, the developers rarely intentionally employ a diverse team of users to test the products before they are shipped. This further reinforces the mistrust for health technologies in the Black community.
5. Lack of culturally competent design team
Most of the AI-powered nutrition coaches available today were developed by tech enthusiasts, rather than people who are culturally competent in food and nutrition. Likewise, the products are rarely subjected to tests on cultural competency.
For context, the performance of an AI model is enhanced by training it on standardized, inclusive data. Standardization helps to prevent some features that present more often from dominating in the model, which can introduce bias in the result. Developers mostly favor standardization—which is easier to achieve—over inclusion. However, it is not always their fault. The omission is sometimes due to a paucity of data from racial minority groups.
Building culturally-relevant AI nutrition coaches
Any AI nutrition coach app that will be relevant to the Black community must be trained using data that incorporates cultural context. The first step is for the developers to work with Black nutritionists who are highly knowledgeable in local cuisines to standardize the preparation and nutritional contents of indigenous dishes. The standardized data should be incorporated into the training of AI-powered nutritional coaches.
The outcome will be the development of culturally aware nutrition algorithms that will make digital health tools more accurate for users in the Black community. Unless this is done, AI nutrition coaches will continue to label native dishes as unhealthy and remove them from its recommendation list.
Until a culturally competent AI nutrition coach is achieved, our community should use these digital tools as supplement rather than substitute, and book regular appointments with certified nutritionists within our community.
For More Reading
- Soul food. https://www.britannica.com/topic/soul-food-cuisine
- Association of Clinical and Social Factors With Excess Hypertension Risk in Black Compared With White US Adults. https://pubmed.ncbi.nlm.nih.gov/30285178/
- How to Create a Healthy Soul Food Plate – Guide and Recipes. https://www.healthline.com/nutrition/healthy-soul-food
- Exploring AI-Based System for African Food Weight-Loss Recommendations. https://www.researchgate.net/publication/385612801_Exploring_AI-Based_System_for_African_Food_Weight-Loss_Recommendations
- 10 Fermented Foods in West Africa & Their Health Benefits. https://mysasun.com/blogs/bloglearning-bytes/10-fermented-foods-in-west-africa-their-health-benefits
- Is Our Soul Food to Blame for Diabetes? https://www.elevateblackhealth.com/is-our-soul-food-to-blame-for-diabetes/
