Stavroula Kousta

Reason and emotion in advocacy

To raise awareness about their causes and ultimately shift public attitudes, advocacy organizations need to stimulate public conversation. Different theories emphasize either rational arguments or emotional appeals as the most successful means through which advocacy organizations can draw people into conversation. New research by Christopher Bail, Taylor Brown and Marcus Mann at Duke University comes to a different conclusion: organizations can best engage the public in conversation if they produce rational arguments after extended emotional deliberation, and vice versa. The authors used text-analysis methods to measure the frequency of emotional and cognitive language use within two advocacy fields, autism spectrum disorders and human organ donation, on Facebook over a period of 1.5 years. They then applied time-series models to the data, finding evidence for ‘cognitive–emotional currents’: an inverse relationship between cognitive and emotional language use over time. Using data collected from 92 relevant advocacy organizations over the same period, the authors show that organizations whose posts contribute to ‘phase shifts’ in the conversation (from emotional to rational debate and vice versa) are more likely to stimulate conversation. This work furthers our understanding of the nature of public deliberation and offers valuable insight for advocacy organizations, which face considerable competition for public attention.

Shifting social norms

The role that institutional decisions play in effecting social change is a question with a long history, but evidence that such decisions play a role in shifting personal attitudes is mixed. Attitudes, however, are not the only determinants of behaviour. Perceptions of social norms — what most others in a group believe or do — are also consequential. In new research, Tankard and Paluck asked whether perceptions of social norms and personal attitudes changed as a result of the June 2015 US Supreme Court ruling in favour of same-sex marriage. In an experiment carried out before the ruling and a five-wave longitudinal time-series study in the few months before and after the ruling, Tankard and Paluck found that perceived norms shifted towards increased support for gay marriage and gay people in response to the actual ruling (time-series study) or a hypothetical favourable ruling (experiment). Although a positive effect on attitudes was only obtained in the experiment, the shift in perceived norms was robust across the two data sets, as well as a third replication data set. The study provides evidence for a substantial effect of a landmark US Supreme Court ruling on public perceptions of support for gay marriage. A crucial outstanding question is how this shift in perceived norms might ultimately affect individual and collective behaviours.

Archaeology: Origin of gender inequalities

Studying dietary change can provide powerful insights into the lifeways and social structures of ancient people and can be especially revealing of power relationships. Yu Dong of Shandong University and colleagues asked whether a shift in subsistence from millets to wheat, barley and animal products during China’s bronze age was associated with societal changes. They performed stable isotope analyses on human bone samples from archaeological sites in China’s Central Plains. They found evidence of a dietary shift from the late Neolithic to the Eastern Zhou dynasty (771–221 bc), which was more pronounced in women: although there was no gender difference in dietary signatures in early farming communities, Eastern Zhou females showed signs of a diet lower in animal products than males, suggesting that meals were no longer shared at the household. The sharp differentiation of male and female diets was associated with increased height disparity between the genders, as well as with a reversal in the distribution of burial wealth: females went from having more burial goods than men during the late Neolithic to having fewer burial goods in the Eastern Zhou. By identifying biological and social correlates of a key dietary shift from the late Neolithic to the bronze age, the authors illuminate the potential origins of malebiased inequality, which persists to the modern day.

Social networks: Inferring financial status

Social ties and economic well-being go hand in hand: more, stronger, or more diverse social connections are associated with better job opportunities, social mobility, and economic development. However, it is still an open question whether social network metrics can be used to infer the financial status of individuals. Research by Shaojun Luo and colleagues at City College of New York and Grandata Labs now leverages two massive datasets for the entire population of a Latin American country to show that there is a near-perfect correlation between social network position (measured through phone data) and financial status (measured via credit card limits). The top 1% and bottom 10% of individuals by wealth show very different communication patterns, which mimic economic inequality at the country level. The researchers identify a metric of network influence —‘collective influence’ — that captures the differences in network structure between the rich and the poor, and enables identification of individual financial status. They then use this metric to target individuals in a social marketing campaign. They find that individuals with high collective influence based on their phone data are much more likely to respond to a marketing campaign for a new credit card than individuals with low collective influence. The ability to infer the financial status of individuals from their social network position provides an important tool for social intervention and marketing campaigns. Future research is needed to determine whether the findings apply beyond the specific population and to examine how the principle of collective influence emerges in the first place.

Reproducibility: Power failure

Replication failure is partly the result of power failure — underpowered studies (coupled with selective publication of only positive results) substantially exaggerate published effect sizes. The nascent field of metascience is playing an important role in quantifying the magnitude of the problem and pointing to solutions. A research paper by Denes Szucs and John P. A. Ioannidis analyses nearly 4,000 cognitive neuroscience and psychology papers published in 18 different journals to estimate effect sizes and statistical power. The authors find that studies in both disciplines are underpowered due to small sample sizes, a problem that is more pronounced for cognitive neuroscience. Low statistical power increases the probability that a study reports false-positive findings. Indeed, Szucs and Ioannidis estimate that more than half of statistically significant findings in psychology and cognitive neuroscience are likely to be false. Finally, they find that higher journal impact factors are associated with lower statistical power. The authors make a number of suggestions for addressing the problem, including compulsory power calculations with minimally required power levels.

Understanding language change

Languages change over time, which makes it harder to trace their history and genealogy further back than 10,000 years. However, it is unclear whether all aspects of language, for instance, grammar and vocabulary, change at the same rate. If not, identifying rates of change for different linguistic features may enable us to reconstruct language history more accurately. In new research, Simon Greenhill and colleagues ask whether different aspects of language evolve in different ways and at different rates. The authors apply Bayesian phylogenetic modelling to grammatical structures and basic vocabulary items from 81 Austronesian languages to infer rates of change. Unlike previous research, they find that grammatical structures changed much more rapidly over time than basic vocabulary. The grammatical features the authors studied were distributed more or less evenly in three rate-of-change categories: fast, medium and slow. Basic vocabulary items, however, nearly all fell into the slow category. The grammatical structures that changed slowly appeared to be consistent with earlier reports and tended to be more abstract than faster-changing grammatical features. This work suggests that looking at language change in a more fine-grained manner may offer more reliable information on which to base inferences about language history and genealogy, extending further into the past.

Unlove the bomb

Interventions in other countries can take a variety of forms: from diplomacy to overwhelming military force. Putting the legitimacy of foreign interventionism aside, a key question is how effective different types of intervention are, especially when loss of civilian life is involved. Melissa Dell and Pablo Querubin examine the effectiveness of US firepower strategies in the Vietnam War. By comparing otherwise similar hamlets that were either air-strike targets or not, they find that bombing had the opposite effect than that intended: it strengthened support for the Viet Cong, increased the likelihood of Viet Cong counter-attacks, and weakened local government. The US Army and the US Marine Corps took different approaches towards counterinsurgency in Vietnam: the army relied on overwhelming firepower, whereas the marines focused on winning hearts and minds. By comparing hamlets commanded by the army with nearby hamlets commanded by the marines, Dell and Querubin find that the marines’ approach of providing development aid and collaborating closely with local security forces was associated with better outcomes and more positive attitudes towards the United States and the South Vietnamese government. The implications of Dell and Querubin’s findings are of current relevance: air-strike precision has improved, but still involves significant loss of civilian life. Their research suggests that politicians and pundits who argue for firepower-based military interventions ought to think again.