Complexity-GAINs international summer school 2022

A socio-ecological system (SES) is a system with consistent, sustained interactions between biological, physical, and social factors. While a relatively new concept, SES has proven helpful in understanding complex systems involving bidirectionality and feedback mechanisms between humans and nature. This is becoming increasingly relevant in the age of climate change and in understanding the impacts of climate-related threats and climate adaptation. We sought to explore the ef fects of climate change on a longstanding, successful SES of Balinese rice-growing communities known as Subak.  There are few SES as successful and consistent as the Subak of Indonesia. For over 1,000 years, they have successfully shared their most valuable and vulnerable resource, water , through investments in steering capacity and collective action. Their stable and time-tested system of cooperation makes them an ideal test subject for exploring how and why SES might fail or adapt to increasing climate volatility.

With Ronan Arthur, Stefan Thurner, John Stephen Lansing

Social media platforms have made politicians change the way they address their audiences as they have become the new dominant media, guiding the topics of interest for politicians and elites (Freelon et al. 2018 ). Traditionally, politicians would set the agenda on issues that are relevant to the public through their discourse. There is a reinforced cycle of opinions where politicians can both follow trends in social media, which could be used as a proxy for priorities of the public, and guide the discourses of the audiences based on their own interests (Barberá et al. 2019 ). However, what is being said in Congress or Parliament may not always match what public figures say online. The Dobbs v. Jackson ruling by the U.S. Supreme Court that overturned the fifty-year old Roe v. Wade decision made headlines internationally, bringing the issue of abortion rights to the surface with politicized opinions against or in favour depending on the party of affiliation, which could be amplified by the Twitter algorithmic personalization as shown before (ibid.). A study showed differences in Twitter engagement by congresspeople (Shapiro et al. 2018 ) using the speech-based categories proposed by (Hemphill et al. 2013 ). Specifically, their results show senators posted more tweets related to positioning statements, e.g. being in favour or against another politician or political issue, while representatives engage more with requesting action, e.g. asking followers to do something online or in person. This research group is particularly interested in understanding the temporal changes on the discourse of congresspeople regarding the leak of the draft version of the mentioned ruling on May 2nd until the actual ruling on June 28th, 2022. Complimentary to such discourse change in Congress, we are interested in exploring the Twitter discussions on the subject. More specifically, this groups research question was: does Congresspeople's issue attention differ on Twitter and in their Congressional speeches? We put forward two hypotheses: 1) Issue attention to reproductive rights would increase during the time of the leak, in Congress and on Twitter and 2) Topic discussion frequency is positively associated with issue attention on Twitter. This second hypothesis is concerned with online engagement measures including likes and retweets (RTs). Furthermore, we would also like to explore the corresponding retweet networks in order to understand public opinion of a highly polarised subject such as reproductive rights for women in the United states. Overall, we want to understand changes in public opinion about reproductive rights in Congress and on Twitter over this period and map the resulting temporal discussion networks.

The existence of online social environments has increased interpersonal connectivity  and  led  to  unprecedented  information access.   Despite  these  avenues for social advances, growing polarization, conflict, and hostility have been increasingly prevalent in the past decade (Yang et al., 2020; Iyengar et al., 2019). A wealth of research has spawned trying to understand the relationship between connectivity and polarization. Network models are a powerful framework to address this issue as they allow us to study phenomena emerging from large systems of interconnected elements.   In  this  paper,  we  start  from  an  important  network  theoretical result  showing  how  minorities’  online  visibility  is  limited  by  two  common principles of human behavior in social environments. Karimi et al. (2018) and Espín-Noboa et al. (2022) study network structures grown from Homophily — our tendency to connect with those similar to  us — and Preferential  attachment (PA) — our  tendency  to  connect  with already popular individuals.  They show that the interaction of these two mechanisms of link formation harm minorities’ connectivity with the majority and consequently their overall visibility in the network.  More specifically they show that node  degree based recommender system will not represent minorities adequately.  Google’s PageRank algorithm and Twitter’s WhoToFollow algorithm are influential examples that suggests information and followers based on node degree (Karimi et al., 2018).  Wang et al. (2022) extends this work by studying the consequences of homophily and PA on simple and complex contagion processes such as information spread.  They demonstrate a complex relation between network structures and contagion properties and a general 'price of fairness':  ensuring information equality comes at the cost of spreading efficiency, and vice versa. In this paper, we study a related but different scenario. Our first research question asks how homophily and PA influence minorities’ ability to overtake the general opinion.  Secondly, we are interested in how minorities’ opinion dynamics change as we introduce multiple conflicting minorities. According to Turchin (2013), conflicts between elite minorities has historically  led  to  social  unrest  by  destabilising  the  ruling  powers.   Troubles between minorities is not limited to the powerful classes.  In any normally distributed  opinion  there  exists  two  tails  of  small  strongly  conflicting  mi- norities.  Secondly,  seemingly united minorities break into conflicting sub- communities such as the division of the Reddit based ‘anti-work’ movement into conflicting ‘work reform’ and ’anti work’ sub-movements. To study this, we extend the work of Espín-Noboa et al. (2022) with the use of an opinion dynamics model,  to see if network structures originated from different homophily and PA settings have an effect on the final representation and distribution of opinions.  For this, the group identity label is set as a fixed attribute to each node, while the opinion on a certain issue is allowed to shift given certain rules: each group has an initial and continuous preference for an opinion, but some individuals may be convinced by their surroundings  to  shift  to  the  other  group’s  opinion.   While  we  expect  the outcome of these simulations to resemble the conclusions Espín-Noboa et al. (2022) extract from the study of rankings, the relation between the two is not trivial, as network structure plays a different role in the performance of opinion dynamics models.  For example, highly segregated networks in high homophily settings may benefit minorities in the rankings (preservation of degree) but would not allow their opinions to gain control of the discussion.

Minority groups are often disadvantaged in social networks.  Previous work has highlighted that minority disadvantage can arise in the presence of two ubiquitous features of social networks: homophily – a preference for connecting with one's own group – and preferential attachment – a preference for connecting with those who are already better-connected [6]. Combined, these two mechanisms can easily result in structural inequalities between groups of different sizes and marginalization of minority groups [2, 6, 9]. In networks, this disadvantage can be captured through the centrality of minority group members, which captures the visibility of minorities and their access to social capital [9]. Previous work has explored how network inequalities emerge from interactions of minority and majority groups of different sizes [2, 6] and how changes in group size and homophily of minority and majority members affect the opportunities of minorities to improve their standing in networks [9]. Yet, little is known about how minority and majority groups fare in the presence of additional groups that could bridge the majority/minority gap, such as majority group members who support the minority group, and minority group members that are incorporated into the majority group. This study attempts to chart the social dynamics of networks that include such groups, and understand under what conditions the presence of such groups can reduce inequity.

Monkeypox is a zoonotic orthopoxvirus first detected in humans in the Democratic Republic of the Congo (DRC), in 1970 (Beer et al. 2019, Bunge et al. 2022). Historically, human cases of monkeypox have been largely limited to the DRC and neighboring countries (Weinstein et al.2005). However, in recent years there have been growing concerns in the research community about the potential for increased spread, particularly due to waning coverage of smallpox vaccination in the wake of smallpox eradication (Weinstein et al. 2005, Bunge et al. 2022, Beeret al. 2019)  as well as increasing encroachment of the built environment and population density (Beer et al. 2019). Supporting this, an increase in incidence and outbreak reports has been observed in the decades since discovery (Beer et al. 2019).In 2003, the first cases of human monkeypox outside of Africa were reported in the UnitedStates. Forty-seven confirmed and probable cases in six states were linked to contact with pet prairie dogs imported from Ghana (Langkop et al. 2003).  In early 2022, Bunge et al. conducted a systematic review of the monkeypox literature and raised the alarm about monkeypox outbreak potential. On May 18, 2022, Monkeypox was again identified in the USA (Kaiser FamilyFoundation, 2022) in what would prove to be the largest outbreak outside of Africa. Swiftly after this diagnosis, researchers ranging from epidemiology to psychology began to investigate the repercussions of monkeypox in the US. A notable characteristic of the virus was that it seemed to be spreading globally among men who have sex with men (Kupferschmidt, 2022, Endo et al. 2022), whereas monkeypox previously had relatively limited spread in African nations (Weinstein et al. 2005). In early summer of 2022, state and local labs began narrow testing to reduce the spread of monkeypox (Doucleff, 2022). Vaccination was largely limited and targeted; for example, in New York the vaccine was initially made available for adult men who have sex with men who have had more than two sexual partners in the past two weeks (NYC Health,2022). Given the limited information available, this targeted intervention was deemed the best way to reduce the spread of this virus. However , men who have sex with men frequently experience stereotyping, stigma, and prejudice on the basis of their sexual orientation.   With this in mind, it is crucial to investigate the efficacy of a tar geted intervention lest this group be exposed to further stigmatization.  We hypothesized that since men who have sex with men sometimes also have sex with women (for example, bisexual men), it could be possible for monkeypox to spread further into heterosexual networks.  To answer this question, we turn to modeling. Grant et al. (2020) presciently predicted the epidemic potential of monkeypox in their 2020 modeling work. Yet, there are few other models available on monkeypox in the literature. Several studies have modeled transmission dynamics of MPX incorporating both human and non-human hosts (Bhunu et al. 201 1; Usman and  Adamu, 2017; Somma et al. 2019; Peter et al. 2021,  Yuan et al. 2022), but the global novelty of the current outbreaks warrants a new approach. Endo et al. (2022) implemented a model of the current outbreak suggesting that a high number of partners could be driving the cases in men who have sex with men. However , their model is limited in that they did not consider women who have sex with women. Here, we implement an agent-based model of monkeypox spread in a sexual network, including hetero-, bi-, and homosexual men and women, and show preliminary results. 

The theory of critical slowing down identifies changes in variance and autocorrelation of a signal as predecessors for critical transitions. More precisely , an increase in variance is said to reflect a stronger response of a system to external disturbances, while a decrease in autocorrelation is said to reflect a longer recovery to a system’ s original state. For instance, the theory of critical slowing down has been used to predict transitions from healthy states to states of mental illness.  We applied the theory of critical slowing down to the prediction of society in crisis reflected by the onset of civil war. Based on the civil war dataset we trained multiple random forest classifiers for the prediction of civil war onset including combinations of raw variables, their variance and autocorrelation as predictors.  A comparison of F1 scores revealed that the inclusion of early warning signals in the analysis did not yield incremental predictive power. A follow up analysis of trajectories of countries experiencing civil war on the first two principal components of all predictor variables however indicated a phase transition happening before the onset of civil war.

publication in progress