Past event
Grand Challenge — Collecting Social Interaction Experiments Online with DuckSoup
GMT (UK)
An update on DuckSoup, the free and open-source platform for running, collecting, and analysing real-time social interaction experiments online — with a live demo and a discussion of scaling social interaction research.
Host
- Jonas Nölle
Speaker
- Pablo Arias Sarah — School of Psychology and Neuroscience, University of Glasgow
“Collecting social interaction experiments online with DuckSoup”
Agenda
- 14:00–14:03 Introduction
- 14:03–14:20 Experimental code, design, and deployment
- 14:20–14:45 Data collection, server monitoring, participant monitoring
- 14:45–15:00 Data analysis examples with open-source tools
- 15:00–15:30 Open discussion
This meeting echoed one of the very first MOSAIC sessions. Pablo Arias Sarah (Marie Curie Fellow, School of Psychology and Neuroscience) presented an update on DuckSoup, a platform to study and experimentally control social interactions in real time online. What seemed like science fiction two years earlier had become a reality, and Pablo gave a demo and led a discussion on the possibilities this new platform opens for multimodal social interaction data collection and research.
Abstract
In 2022, at a MOSAIC Grand Challenge, I presented why it is of interest to use voice/face transformation algorithms to transform participants’ voice and face in real time during free social interactions — giving researchers the ability to make causal inferences between transformed attributes (e.g. vocal gender, smile intensity) and subsequent behaviours (e.g. social outcomes, expressive synchrony).
This year I introduce the experimental platform DuckSoup, a free and open-source platform we developed to this end. I provide a demo of how to use DuckSoup alongside other free and open-source technologies to run, collect, and analyse social interaction experiments online: how to design an interactive experiment using the Python package oTree, how to deploy it online, how to run it with real participants, and finally how to analyse video recordings to extract vocal, facial, and semantic information. We close with a discussion of the benefits of scaling social interaction research, the possibilities opened by novel video-analysis pipelines, and how these could help address key theoretical questions.