Economic Agent-Based Models: Crossing Over the Tipping Point

Economic agent-based models are half a century old, but have remained mostly at the margin of economics. In recent years, a few successes have raised hopes that we are approaching a tipping point, where ABMs’ superior performance will attract more researchers, feeding up further performance increases.

In recent years, SFI has primarily organized short workshops, working groups and meetings – only lasting between two and five days. Our thinking has been that that length balanced the cost to the Institute with the benefit to the researchers. Researchers used their time at SFI to share insights and brainstorm new approaches. The bulk of the new research is then done after researchers return to their home institutions. However, we believe that a prolonged gathering of researchers offers the opportunity to drive more substantial progress in a particular area.

This month-long working group will generally focus on Agent-based Modeling (ABM) Specifically, this project will aim to “fill-in” intellectual gaps needed to enable the broader adoption of this approach within academia, industry, and government. Although there has been a wealth of research on ABMs, we believe a few key pieces of work could push this field over a tipping point of broad acceptance and application. Once that threshold is crossed, we believe more scholars will be interested in refining the core techniques, and practitioners (leaders from industry, government, and civil society) will be more willing to use ABMs in their analysis and policy assessments.

The group has identified four component areas in which they feel modest progress could stimulate significant additional work on agent-based models:

(i) Technical challenges related to data ingestion.
(ii) “Process validity” for individual behaviors.
(iii) Benchmarking in specific domains.
(iv) Agent-based approaches to modeling climate change and supply chains.

This meeting was made possible by generous support from the Zegar Family Foundation and the Omidyar Network.