[Research Seminar] Nego/Sales/Pon ICON: “Conversational receptiveness: Improving engagement with opposing views” M. YEOMANS – Imperial College London
Speaker: Michael YEOMANS
Assistant Professor in Strategy & Organisational Behaviour – Imperial College London
Co-authored with Julia MINSON, Hanne COLLINS, Francesca GINO, & Burint BEVIS
Date and Location – Friday April 15th 2022 from 14:00 to 15:30
in P400 (Paris Campus) and B252 (Lille campus – On Zoom)
—————————————–
ABSTRACT
Disagreement is unavoidable in social life but can also give rise to biased inferences and conflict. We build a domain-general and interpretable machine learning algorithm for “conversational receptiveness” – language that communicates thoughtful engagement during disagreement. We first pool data from crowdsourced workers arguing about controversial topics (N = 2,860; preregistered) to build a model using the politeness R package. This model identifies structural features such as acknowledgement (“I understand”), subjectivity (“I believe”), and hedges (“sometimes”) that demonstrate listening and intellectual humility – as well as some features that do the opposite.
We show that people who express receptiveness during disagreements are more persuasive, build trust, and prevent conflict escalation, and we show two pathways to increase receptiveness in others. We first show social contagion – expressing receptiveness encourages disagreeing others to be more receptive – in four samples: lab participants, Wikipedia editors, MOOC students, and local government officials.
We next showed that our algorithm is interpretable – both to researchers and to users. We compare two interventions – a static “recipe” describing the main features and an algorithmic coaching system that provides immediate, personalized feedback on participants’ feature use (N = 1,104; preregistered). Both interventions improved receptiveness over baseline, but while the recipe’s effects declined over time, the feedback system’s effects increased over time.
Overall, we demonstrate that conversational choices have important interpersonal consequences, and these choices can be encouraged using interpretable algorithms.