Science Communication in Light of Moral and Social Concerns: Testing a Statistical Physics Model of Belief Change

Abstract

Most efforts to increase public acceptance of scientific facts have focused on providing transparent factual information. While this is essential for enabling people to make informed decisions, evidence is mixed about the effectiveness of facts alone for modifying beliefs. To enable more effective science communication, we need a better understanding of how moral and social considerations affect belief change beyond facts. In two longitudinal studies with a total of 1,673 participants, we experiment with different educational interventions that present scientific messages about the safety of GM food and childhood vaccines designed to alleviate different moral and social concerns. To aid our understanding of why some interventions are more successful than others, and why only some people change their beliefs, we develop a statistical-physics-inspired quantitative model of belief dynamics. The model incorporates several possible mechanisms of belief change and can provide insights into the cognitive processes underlying successful interventions and individual belief change. Our experimental data support the theory-driven hypothesis that belief change is more likely when facts are presented in a way that lowers their dissonance with people’s moral and social concerns, but only when the particular type of dissonance (moral or social) is considered important. In addition, individuals who pay more attention to how facts relate to their moral and social concerns show more belief change with decreasing dissonance. Taken together, our model and results provide a pathway towards a more rigorous understanding of how to address people’s moral and social concerns when communicating scientific facts.

Integrating social and cognitive aspects of belief dynamics: towards a unifying framework

“…We provide two empirical examples. One example (figure 2 b ) involves beliefs about safety of GM food, empirically measured two times in each of the three waves of a longitudinal survey on a national sample of N = 531 participants [161]. Besides this focal belief, in each wave, participants also reported 12 other beliefs about their individual moral issues related to GM food [162], and perceived beliefs of different social groups (family, online community, scientists, doctors, etc.).…”

“…Parameters w and β are fitted with a grid search procedure (ranges: 0 ≤ w ≤ 1, and 0 ≤ β ≤ 100) for each time point on a random half of the participants, so that they minimize the deviation between modelled and actually measured focal beliefs (inspection of the deviations showed that only one parameter combination minimized the deviation between model and data). Here we present belief trends in the group of participants who received only the basic scientific information about the safety of GM food (figure 2b; for further results, see [161]). On average across time points, parameters w and β for this group are estimated at 0.57 and 23, respectively, suggesting a stronger weight on social dissonance and a moderate amount of uncertainty in belief updating process.…”

“…Across 50 random half-splits, we find that our models predict this unseen data well (figure 2 b , right panel). When comparing models assuming different integration rules, we find that this group of participants do not seem to consider all potentially relevant moral and social beliefs, but seem to rely on one most important belief, as reflected in lower root-mean-squared deviations and higher correlations of predicted trends for the most important rule (0.07 and 0.85, respectively) compared to the simple average rule (0.12 and 0.58) and weighted average rule (0.12 and 0.60; see [161] for more details and further insights from this study).…”

Using a cognitive network model of moral and social beliefs to explain belief change

“…These models tend to consider either the dissonance between moral and social beliefs using the concept of energy from statistical physics without considering the relationships between these beliefs (34), or the network imbalance between personal and social beliefs (e.g., belief A is positively connected to beliefs B and C, but beliefs B and C are negatively connected) (35)(36)(37). Dissonance-based models can explain belief change using estimated energies from reported moral and social beliefs (38). However, these models do not take into account the network structure of interrelated moral and social beliefs.…”

Beyond collective intelligence: Collective adaptation

“…Fortunately, scientists have never had access to so much empirical data about different aspects of human cognition and sociality as today. It is now more convenient than ever to obtain data from group experiments and observational studies [ 266 ], historical and archaeological data, ethnographic data, mobile phone data [ 267 ] and other sensing methods [ 268 ], to longitudinal surveys [ 269 , 270 ], and analyses of large textual corpora and other by-products of human interactions (e.g. [ 271 , 272 ]).…”