Ceres2030 researchers meet at the FAO in Rome, June 2019. Photo: Trevor Butterworth, Ceres2030.

While the gains in in agriculture over the past 60 years and the reduction in global hunger are extraordinary, we face problems both old and new and equally pressing. One in nine people are still hungry. Agriculture presses on fragile environments, threatening biological diversity and the availability of clean freshwater. Simply put, agriculture is complex. Food systems—how we feed people—are complex.

But as problems change and evolve, so does knowledge. Agriculture has seen the total volume of research double in the past ten years, and this trend is mirrored across science. The question is how to synthesize this. How do we combine research findings from so many different fields to address food security? Reports and studies  contain hundreds of possibly useful ideas—interventions that could be funded to improve crop storage, livestock feed, conserve water, and preserve land.

A large review, such as the Intergovernmental Panel for Climate Change (IPCC) is one model for how to gauge the quality and comprehensiveness of the evidence on a topic. But synthesizing relevant information is a time-consuming task; the  IPCC reports take years to generate, and we need more evidence today on “what works” for sustainable approaches to ending hunger.

This is why we turned to machine-learning. This gave us the ability to build search tools that could synthesize the ever-expanding volume of agricultural research for policy-relevant interventions. To date, we’ve analyzed more than half-million articles; and now, after an exhaustive evaluative process, more than 70 brilliant and talented researchers from around the globe—all working on a voluntary basis—are examining eight intervention areas that hold promise for the future.

Their findings, subject to peer review, will be published in Nature Research Journals in 2020. The next step will be to use that data to determine how much these interventions will cost to implement. This is vital information for policy makers in the quest to fulfill the mandate of SDG 2, “Zero Hunger” by 2030.

The following sections explain the eight intervention questions, how they were selected, how machine learning can unlock agricultural research for policy makers, and how we are making these tools available to other researchers.

How we assess the quality of evidence