Our Projects
Synthetic Replication Games
The Synthetic Replication Games are a series of one-day events organised in the context of a large-scale replication project exploring the potential of large language models (LLMs) to augment human samples in social science experiments. During each event, researchers of all backgrounds and career stages are invited to collaborate in adapting experimental studies from top-tier journals into a standardised format and replicate them using LLM-powered synthetic participants. The material generated during each event will be consolidated in a battery of synthetic replication packages, which will serve three main purposes: (1) benchmarking the ability of LLMs to simulate human participants in social science experiments; (2) facilitating the use of experimental data to fine tune LLMs; and (3) developing a standardised format for the replication material of synthetic experiments.
Artificially Intelligent Opinion Polling
Roberto Cerina and Ray Duch
The Talking to Machines initiative aims to revist how we conduct large scale experimental research. We believe that many of the key elements of experimental design and implementation can be enhanced by AI. Measuring the attitudes and behaviors of participants in these studies is a typical outcome variable in randomized control trials. An on-going research project explores how AI can radically change public-opinion research by providing researchers with a general methodology to make representative inference from cheap, high-frequency, highly unrepresentative samples.
Synthetic Randomized Control Trials
The Talking to Machines project is exploring how to leverage AI to enhance large scale randomized control trials. Randomized Controlled Trials (RCTs), particularly in low- and middle-income countries (LMICs), have become a staple of policy design and evaluation. The policy relevance of these RCTs is very much determined by our ability to scale-up results and generalize – but generating samples that reflect the characteristics and diversity of populations of interest can be extremely expensive. LLMs may provide an option to alleviate some of these problems. T2M is conducting on-going research to develop a platform to facilitate these experiments and also to assess the feasibility of AI enhanced RCTs.