Reducing socio-ecological conflict using social influence modelling

Abstract

Polarisation of opinions across communities can lead to social conflict, reputational damage and the disruption of operations and markets. Social influence models have been widely used to better understand processes driving conflict from a theoretical perspective. Using aquaculture as a case study, we demonstrate how such models can be extended to accurately hindcast the transition from population consensus to high conflict, including observed catastrophic tipping points. We then use the model to quantitatively evaluate strategies aimed at reducing aquaculture conflict. We found that persuasive advocacy was ineffective and often counterproductive, whereas meaningful engagement, collaborative learning and improving scientific literacy targeted broadly across the population was effective in moderating opinions and reducing conflict. When such messaging was targeted too narrowly or too infrequently, it tended to be negated by ongoing exchange of misinformation within the population. Both the modelling approach and lessons on effective communication strategies are relevant to a broad range of environmental conflicts.

Figure 1

(a) Basic network structure of the Social Influence and Events Model (SIEM). The opinion of an individual can potentially be influenced by media or other individuals with higher average certainty (larger icons) and opinions that fall within their confidence threshold (separation of icons). Because homophily is a system characteristic, confidence threshold is the same for all individuals, whereas opinion and certainty vary between individuals. (b) Processes simulated in SIEM on a monthly timestep.

Figure 2

Time series over the historical period (January 1996 to March 2021) and a projection period based on repetition of the previous 3-years (April 2021 to March 2032). (a) Newspaper framing per article (grey crosses) and annual means (black points). (b) Modelled opinions averaged across the population (grey points) and averaged across the entire ensemble (black line). Average opinions estimated from five community surveys with 95% confidence intervals are also shown for comparison (red points). (c) Modelled population conflict levels (grey points) and ensemble mean conflict level (black line). Average conflict estimated from the five community surveys are also shown for comparison (red points). While no survey data was available prior to 2015, the industry has been described as having ‘an enviable social licence to operate’ over that period, consistent with positive opinions and low conflict.

Figure 3

Movement of average opinion and conflict in response to a range of communication strategies applied from 2021. The orange line indicates baseline values for opinion and conflict when no strategy was applied (case shown in Fig. 2). Examples of time-series plots for collaborative learning and positive persuasion are shown in Fig. S5.