A Digital Dietician for Every Indian
The National Institute of Nutrition (NIN), based in Hyderabad, recently announced its plan to develop an AI-powered platform to help Indians make healthier food choices. The goal is ambitious and timely. With rising rates of non-communicable diseases
like diabetes and hypertension linked to diet, and an increasing consumption of packaged foods, the need for accessible nutritional guidance is urgent. The proposed tool aims to be a one-stop source where users can look up a packaged food product and get a simple breakdown of its nutritional profile. The project involves a partnership with a Hyderabad-based startup to create a scientific database of the thousands of products on Indian shelves.
The Challenge: A One-Size-Fits-All Algorithm?
While the initial focus is on packaged foods, the larger questions arise when considering the full spectrum of the Indian diet: home-cooked meals. This is where the complexity begins, and where critics and researchers are raising flags. The core problem is that AI is only as good as the data it's trained on. For an AI to be effective, it needs a massive, detailed, and accurate database. When it comes to Indian food, creating such a database is a monumental task.
Why Indian Food Is So Hard to Quantify
Most existing nutrition apps were built for Western diets, where meals are often composed of discrete, standardized items. Indian food defies this structure. A single dish, like sambar, can have countless variations in ingredients and preparation from one state, community, or even household to another. The nutritional value of a 'dal' can vary dramatically based on the type of lentil, the amount and type of oil or ghee used, and the specific tempering (tadka). Portion sizes are another hurdle. We don't typically measure in grams; we use intuitive measures like a 'katori' of sabzi or a couple of phulkas, which are not standard units. Researchers at institutions like IIIT-Hyderabad have highlighted that AI struggles to visually distinguish between similar-looking curries in a thali or estimate quantities when foods are mixed, like dal poured over rice.
More Than Just Calories
Even with perfect data, nutrition isn't just a numbers game of calories and macronutrients. Food in India is deeply intertwined with culture, season, and traditional knowledge systems like Ayurveda. Some emerging nutrition platforms are trying to incorporate this 'Ayurvedic intelligence', analysing food based on properties like taste (rasa) and its effect on an individual's constitution (prakriti). These systems recognize that a food's impact goes beyond its chemical composition. The NIN's standard food composition tables are a vital resource, but they can't capture the full context of how food is consumed and perceived across the country.
The Path Forward: Human-Centred AI
The concerns around NIN's AI bot aren't a rejection of technology, but a call for it to be more thoughtful and context-aware. The development of open-access resources like the Indian Nutrient Databank (INDB) is a crucial step, providing detailed data on hundreds of common recipes. However, for any AI tool to succeed, it must be flexible. It needs to accommodate regional variations, understand colloquial food names, and perhaps even learn an individual's specific eating habits over time. The danger of a poorly implemented AI is not just that it's unhelpful, but that it could give misleading or even harmful advice, as has been seen with generic AI chatbots offering diet plans. A tool that can't tell the difference between a home-cooked rajma and a restaurant version isn't just inaccurate; it's culturally illiterate.
