Charlotte O. Brand’s research while affiliated with The University of Sheffield and other places

Recent research on argumentative dialogues has focused on persuading people to take some action, changing their stance on the topic of discussion, or winning debates. In this work, we focus on argumentative dialogues that aim to open up (rather than change) people's minds to help them become more understanding to views that are unfamiliar or in opposition to their own convictions. To this end, we present a dataset of 183 argumentative dialogues about 3 controversial topics: veganism, Brexit and COVID-19 vaccination. The dialogues were collected using the Wizard of Oz approach, where wizards leverage a knowledge-base of arguments to converse with participants. Open-mindedness is measured before and after engaging in the dialogue using a questionnaire from the psychology literature, and success of the dialogue is measured as the change in the participant's stance towards those who hold opinions different to theirs. We evaluate two dialogue models: a Wikipedia-based and an argument-based model. We show that while both models perform closely in terms of opening up minds, the argument-based model is significantly better on other dialogue properties such as engagement and clarity.

Recently, Altay et al. (Altay et al. 2021. J. Exp.Psychol.: Appl. (doi:10.1037/xap0000400)) showed that 5 min of interaction with a chatbot led to increases in positive COVID-19 vaccination attitudes and intentions in a French population. Here we replicate this effect in a vaccine-hesitant, UK-based population. We attempt to isolate what made the chatbot condition effective by controlling the amount of information provided, the trustworthiness of the information and the level of interactivity. Like Altay et al., our experiment allowed participants to navigate a branching dialogue by choosing questions of interest about COVID-19 vaccines. Our control condition used the same questions and answers but removed participant choice by presenting the dialogues at random. Importantly, we also targeted those who were either against or neutral towards COVID-19 vaccinations to begin with, screening-out those with already positive attitudes. Replicating Altay et al., we found a similar size increase in positive attitudes towards vaccination, and in intention to get vaccinated. Unlike Altay et al., we found no difference between our two conditions: choosing the questions did not increase vaccine attitudes or intentions any more than our control condition. These results suggest that the attitudes of the vaccine hesitant are modifiable with exposure to in-depth, trustworthy and engaging dialogues.

Prestige-biased social learning (henceforth “prestige-bias”) occurs when individuals predominantly choose to learn from a prestigious member of their group, i.e. someone who has gained attention, respect and admiration for their success in some domain. Prestige-bias is proposed as an adaptive social-learning strategy as it provides a short-cut to identifying successful group members, without having to assess each person’s success individually. Previous work has documented prestige-bias and verified that it is used adaptively. However, the domain-specificity and generality of prestige-bias has not yet been explicitly addressed experimentally. By domain-specific prestige-bias we mean that individuals choose to learn from a prestigious model only within the domain of expertise in which the model acquired their prestige. By domain-general prestige-bias we mean that individuals choose to learn from prestigious models in general, regardless of the domain in which their prestige was earned. To distinguish between domain specific and domain general prestige we ran an online experiment (n = 397) in which participants could copy each other to score points on a general-knowledge quiz with varying topics (domains). Prestige in our task was an emergent property of participants’ copying behaviour. We found participants overwhelmingly preferred domain-specific (same topic) prestige cues to domain-general (across topic) prestige cues. However, when only domain-general or cross-domain (different topic) cues were available, participants overwhelmingly favoured domain-general cues. Finally, when given the choice between cross-domain prestige cues and randomly generated Player IDs, participants favoured cross-domain prestige cues. These results suggest participants were sensitive to the source of prestige, and that they preferred domain-specific cues even though these cues were based on fewer samples (being calculated from one topic) than the domain-general cues (being calculated from all topics). We suggest that the extent to which people employ a domain-specific or domain-general prestige-bias may depend on their experience and understanding of the relationships between domains.