Most of the empirical work I do with collaborators needs infrastructure that doesn’t really exist off the shelf. Running a single behavioral experiment with a handful of people is well supported; running thousands of them, each with groups of participants interacting in real time, is not.

Lead by Abdullah Almaatouq, and with Joshua Becker, James Houghton, Nicolas Paton, and Duncan Watts, we built and extended Empirica, an open-source virtual lab designed for exactly this kind of high-throughput, macro-level experimentation. Empirica handles the hard parts — synchronization, treatment randomization, real-time interaction, and iterative design — so researchers can focus on the experiment itself BRM’21.

Empirica was one piece of a broader conversation, across many labs, about what it would take to actually scale up behavioral and social science. In a working paper with a large group of collaborators we laid out a vision for shared infrastructure, shared protocols, and shared participant access as the missing middle of the field OSF’21.

This infrastructure turned out to be essential to the work on integrative experiments that came next — you can’t run a design space full of experiments if each one costs you a month of engineering.