5.4 What it takes to do this well

Key points

Each step is hard on its own; doing all seven consistently, across many farms and questions, is the mountain.
Advice that’s right for the wrong reason can cost a season.
A nutrition claim built on a correlation can mislead a whole market.
Causal inference earns the right to say what is likely to happen, for farms like yours.

Chapters

Key terms

Causal inference: The discipline of telling a cause from a pattern: working out what would actually happen if you changed something, and how sure you can be. It combines a precise question, a causal map of the assumptions, a study design that can answer the question, and checks that try to break the answer.
Causal map (DAG): A diagram that draws each thing that varies as a box and each direct effect as an arrow, with no numbers. Also called a directed acyclic graph: arrows point one way and no route returns to its start. Sewall Wright’s path diagrams of 1921 were the first. A map puts assumptions where experts can check them.
Sensitivity analysis: Asking how strong an unmeasured confounder would have to be to overturn a result, and comparing that with confounders you did measure. Jerome Cornfield’s 1959 answer to Fisher about smoking was the first. If the hidden factor would need to be far stronger than soil type, you can be fairly confident.
Proportionate next step: The smallest step that can answer the question. Not every promising pattern calls for a field trial; many questions need only one or two steps, such as completing the records, measuring what’s missing or learning from changes already happening. A trial, which asks the most of a farm, is chosen when the evidence points to it.

Transcript

Last time's quick check: in branch one the link runs through the grain's structure, not the soil, so changing what happens in the field wouldn't change the zinc. This series began with an AI that was right for the wrong reason. It read the hospital, not the lungs. Twenty-nine videos later, here's what it takes to be right for the right reason.

First, a precise question: which treatment, which outcome, for whom, over what time. Lind's question: what exactly is "it"? Second, the evidence that's already published, read critically and weighed by how each study was designed. Third, data labelled honestly: what was measured, what was only recorded, what's illustrative and what's missing.

Fourth, a causal map, drawn before the analysis and checked by people who know the farm, the animals and the food. Fifth, the right design for the question: a trial, a natural experiment, a matched comparison or an emulated trial. Sixth, methods that fit the data, from simple comparisons to machine learning, with their assumptions stated out loud. And seventh, checks that try to break the answer, and the discipline to say "not yet". Each step is hard on its own. Doing all seven, consistently, across many farms, crops and questions, while keeping the evidence and the data straight: that's the mountain.

Why go to all this trouble? Because decisions in farming and food move money, land and people's health. Advice that's right for the wrong reason can cost a season. A nutrition claim built on a correlation can mislead a whole market. Causal inference is how we earn the right to say: do this, and here's what is likely to happen, for farms like yours.

That's the work of Vitagri Vitals, Vitagri's causal engine for agriculture and food nutrition. It brings published evidence, farm and food data and causal maps together, chooses the proportionate next step, and says plainly when the evidence isn't there yet.

Questions

Why does causal inference matter in farming and food?

Because decisions in farming and food move money, land and people’s health. Advice that’s right for the wrong reason can cost a season, and a nutrition claim built on a correlation can mislead a whole market. Causal inference is how you earn the right to say: do this, and here’s what is likely to happen.

What does Vitagri Vitals do?

Vitagri Vitals is Vitagri’s causal engine for agriculture and food nutrition. It brings published evidence, farm and food data and causal maps together, chooses the proportionate next step, and says plainly when the evidence isn’t there yet. The series explains the method, not how Vitals works inside.

Sources

Zech, J.R., Badgeley, M.A., Liu, M. et al. (2018). Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: a cross-sectional study. PLOS Medicine, 15(11), e1002683. doi:10.1371/journal.pmed.1002683

All 30 films · Glossary · The Vitagri Academy