An AI project asks what helped ancient societies endure famine

A team led by Timothy Harrison at the University of Chicago has received $800,000 from Schmidt Sciences for Modeling Famine Resilience, announced on 14 October. The three-year programme will use AI to investigate how societies responded to food crises.[1]

One investigator, Cornell’s Sturt Manning, has already studied the difficulty of connecting weather to history. His team’s 2023 Nature paper used juniper ring widths and carbon isotopes from central Anatolia to identify severe dryness around 1198–1196 BC, with dating uncertainty of roughly three years. The interval coincides with the end of the Hittite Empire, but the authors explicitly lacked evidence to establish direct causation.[2][1]

The programme will align cuneiform and Chinese accounts with archaeological settlement evidence and climate records. AI will search for crop failures, relief and institutional responses; specialists will assess its hypotheses. Plans include new tree-ring records from Anatolia and the northern Levant around upheavals roughly 4,200 and 3,200 years ago.[1]

The accepted hypotheses will enter simulations of households and institutions making decisions under stress. Those results should then generate fresh questions for the AI to investigate: a loop between documentary evidence and models of social behaviour.[1]

The team plans open datasets, software and a public portal for exploring hypotheses and simulation results.[1]

That social layer matters. Manning’s earlier paper describes societies accustomed to bad years through food storage, water infrastructure and social networks. A prolonged crisis can exhaust such arrangements, while neighbouring places follow different trajectories. Finding a drought and finding a collapse therefore leaves a historical task: explaining the decisions and constraints between them.[2]

Report an error or suggest a correction