The Role of Data Science in Food and Nutrition Research
The Role of Data Science in Food and Nutrition Research
It's all about externalities
This week, I participated in a National Academies workshop on the role of advanced computation, predictive technologies, and big data analytics related to food and nutrition research. It was a fascinating couple of days filled with great talks. You can see the agenda here. A proceeding based on the event will be prepared by an independent rapporteur, and slides and video will be posted soon.
The (abbreviated) goals of the workshop were to:
Explore opportunities and challenges of applying data science in food systems and nutrition research.
Discuss appropriate use of evidence from data science to inform food- and nutrition-related programs and policies.
Consider ethical challenges associated with data science in food and nutrition research.
Identify opportunities and challenges for capacity building and training to support the application of data science in food and nutrition research.
(Following the lead of some presenters, I shortened "advanced computation, predictive technologies, and big data analytics" to "data science".)
I used my time to tout the promise of AI and data science to help solve the big challenges in food and agriculture. Here's the 5-minute version of version of my 30-minute talk (slides here).
American farmers now produce more than four times as much they did 100 years ago with only a small increase in inputs. They use more capital and purchased material inputs, but much less labor and a little less land (see Julian Alston and Phil Pardey's work for details). Globally, farmers produce 250% more cereal grains than in 1960 with just 15% more land. These productivity increases are truly remarkable.
Because of increasing productivity, the inflation-adjusted prices of major agricultural commodities are down at least 50% since 1960 (depending on the year). On average, Americans now spend 11% of their disposable income on food, down from 17% in 1960 and over 40% in 1900. The percent of people undernourished in developing countries has declined from 30% to 12% in the last 50 years.
This all seems like a huge triumph. However, food companies have taken these cheap ingredients and created tasty processed foods that we can't resist. As a result, we eat more. Americans now consume 30% more calories than in 1960. Nearly 1 in 4 people in the OECD is considered obese and 8.4% of the health budget in OECD countries is projected to be spent on the health consequences of overweight in the next 30 years. Moreover, modern agriculture causes nitrate pollution of waterways and significant greenhouse gas emissions, among other environmental challenges.
What went wrong?
Answer: Externalities.
An externality is a side effect or consequence of an economic decision that affects other parties without this being reflected in the cost. When a decision maker doesn't see the true cost or benefits of their decision, then they make a decision that is suboptimal. For example, a farmer doesn't face the cost of nitrate pollution of waterways, which may cause them to use too much fertilizer.
For an example integral to this workshop, people make eating choices today that will affect the health of their future self. The current self and the future self are often not tightly connected. When you grab some food on the run or eat subconsciously while relaxing on the couch, you don't immediately feel the health effects. You are not thinking much of how it will affect the health of your future self. (Some people refer to this case as an internality rather than an externality.)
This framing departs from the notion of people as rational consumers making optimal decisions about everything they eat. The existence of the diet industry is strong evidence against the rational consumer hypothesis. Changing people's food decisions to be sufficiently healthy seems like an intractable problem, at least with the economist's tools of taxes and subsidies.
Pharmaceutical and surgical solutions may be a partial solution. New weight-loss drugs such as Ozempic slow digestion and suppress appetite (similar to bariatric surgery). In their current form, these drugs require daily injections for the rest of the patients life, and they are very expensive. They are not a population-level solution at present.
Another approach is to develop better foods. What if we had foods that are tasty, convenient, and healthy? Why do we have to choose between health and convenience? Why do we have to choose between health and taste?
This is where AI can come in, both in mapping foods to health outcomes so we understand better what healthy means and then developing foods with healthy traits. (The two links in the previous sentence go to AIFS researchers, but researchers at the workshop presented many other examples.)
As I wrote in Ag Data News last month, solving the big challenges facing agriculture and food requires a lot of things the market won't demand. In economics jargon, we can't expect the private sector to solve the externality problem but we can invest in technological "why not both?" solutions.
Postscript: I was listed in the workshop agenda as talking about farmer's trust in AI. I addressed that topic by noting that farmers want solutions that solve the problems they have. This means that AI tools that may improve environmental outcomes (e.g., robotic weeding, precision irrigation)) must also solve problems like regulatory compliance, labor shortages, and tight profit margins. Otherwise farmers will not use them. As with food consumption, the technological solution to the externality is one that solves both the external and the internal problem. Why not both?