Talita Greyling (p)

Annie Tubadji

Unfortunately, despite research showing that receiving the COVID-19 vaccine is the best way to protect yourself, your loved ones, and your community against contracting the virus, vaccination rates in the Western world are slowing down, and there is a sense of increased complacency. We now know that negative emotions such as fear related to, for example, side effects influence peoples' attitudes towards receiving the vaccine. We also know that happier people make better health-related decisions since happier people are less inclined to engage in high-risk activities such as smoking. Given the aforementioned, the primary aim of this paper is to determine those factors most important for achieving higher levels of vaccination rates. We employ multiple supervised machine learning algorithms to achieve this aim. In our analyses, we include country-level factors for ten countries in Europe, Africa and Australasia. These factors include happiness, collective emotions, economic and socio-economic features, COVID-19-related data, policies, and trust in institutions. To measure happiness, collective emotions and trust in vaccines and institutions, we derive time-series data from the Gross National Happiness.today project constructed using Big Data and Natural Language Processing techniques. Our dataset, which includes high-frequency daily data, is unique and has the advantage of being timeous. Our findings provide actionable policy insights which can potentially increase vaccine uptake.

Frederic Boy (p)

Talita Greyling

The rapid uptake of digital technologies across human life generates an unprecedented volume of user-created data that is forecast to reach 4.63*1020 bytes by 2025. This sheer amount of data creation provides an opportunity to understand peoples’ informational needs and ultimately develop tools that better serve individuals, organisations, and society. To do this, the methods of Artificial Intelligence (AI) must be better integrated within the social sciences questioning of contemporary realities. Doing so will promote greater uptake and provide optimal insights to the public, policymakers, and journalists.
In the current environment, digital technologies mediate our relationships and experiences. People rely on digital services and devices to consume, communicate, be informed, and be entertained. Finely grained insights from the interplay between technology usage and people’s psychological, social, economic, political, and cultural digital experiences are essential to understand contemporary society. When equipped with cross-disciplinary knowledge, at the intersection of social & behavioural Sciences and AI, researchers and practitioners will be better able to map the functional connections between sections of individuals’ digital lives and the zeitgeist, the defining spirit or mood of an epoch.
The dynamic informational contexts produced by the COVID-19 pandemic and the current Russian aggression on Ukraine are unprecedented and unanticipated opportunities to understand how sudden global shocks modulate people’s online searches.
The present paper presents a series of historical and real-time analyses investigating how highly granular digital data can augment population-scale knowledge gained from traditional means. I will present how we validated strong temporal linkages in the digital surveillance of search engines' time series during COVID-19 and the conflict in Ukraine. Then I will show how high-frequency search-listening analytics provide robust, finely-grained, and replicable evidence on variation in population-level-aggregates of mental health. Finally, I will review the evidence we gathered in multilinguistic search-listening research that analysed the relationship between online search behaviour and an individual’s well-being in Romania, France, Turkey, Italy, Germany, Ukraine, and the United Kingdom.
This body of evidence will be discussed against the backdrops of 1-a global information war where facts are targets and the truth not absolute, and 2-the need for academic research to transparently contribute to the intelligent systems that analyse, synthesize, and safeguard data. Understanding the informational needs that characterise an epoch can help expose and counter the information distortions onto which autocrats prosper and rewrite historical narratives.

Annie Tubadji, Yashi Jain (p), Talita Greyling , Stephanie Roussouw

Frederic Boy

Vaccines are clearly a means to protect the life of the vaccinated and indirectly the lives of the other members of society hence the decision to get vaccinated is a proxy for individual concern with social welfare. This study asks whether the behaviour under a shock condition, such as the vaccination for COVID-19, can be predicted by other social welfare-relevant behaviours in a locality – such as the ecological concerns of individuals living in a certain locality.
Using big data from individual Tweets about vaccines (representative on the province level) and actual vaccination behaviour on the provincial level for England and Wales in 2019-2021) we implement sentiment analysis using AI algorithms to identify the pro- or anti-vax sentiment of each Tweet. Next, we compare the expressive rhetoric in Tweeting (which clearly does not have any direct practical effect on contagion) and actual vaccination behaviour when the cost of lives is importantly factored in the behaviour. Finally, we obtain instrumental variables – proxies for ecologically relevant behaviour such as the use of cars and the sorting of garbage in a living place before the pandemic, and we use these proxies for the context of social welfare concern as a predictor of the individual emotion and preference towards vaccination. We employ the Culture Based development approach to quantify local cultural context and the stock of local cultural capital to further delineate between the cost-benefit of vaccination and the clearly identified cultural bias on the regional level. A hierarchical model shows clearly the statistical sources of influence on the final individual preference for or against vaccination.
Our study helps to distinguish empirically the rational cost-benefit analysis of vaccination from the cultural impact of the local social welfare concern. These findings are particularly helpful because they indicate how seeming bounded rationality emerges not due to cognitive boundedness, but due to cultural embeddedness and social pleasing of the context in which an individual finds themselves embedded.

Annie Tubadji (p), Haoran Huang, Thora Tenbrink, Mat Comfort

Yashi Jain

A core toy model in Culture Based Development (CBD) is the model of cultural valuation of economic assets. Previous CBD research has provided evidence for the cultural valuation anomalies under the information treatment that a given asset is produced by an AI – namely people tend to under-evaluate music when they learn it is composed by AI and to upgrade their evaluations for the human compositions (see, Tubadji, Huang and Weber and then Tubadji (2021) and Tubadji and Huang (2023)). The current study aims to first replicate the above-described CBD experiment, this time using a different form of art – poetry (in its short form – haiku), composed respectively by human poets and by different versions of Chat GTP algorithms. Second, the current study extends the inquiry by posing the research question – are people’s emotions, perceptions and awareness about the ecology and the world more readily impacted by the huma creative art product or by the AI generated art. There are reasons to believe that either is potentially plausible. As machines do not cheat, they are more trustworthy so they may be perceived as more worth following one’s recommendation and advice. However, the literature shows that people prefer human-error exhibiting car navigation systems than more efficient GPS systems. Our analysis equips us with further insights on the potential for impact by ChatCPT and related AI technologies on the evolution of the socio-economic discourse. Potential reasons for our results stemming from the very language optimization procedures that are fundamental in information science are offered as speculations, which in light of our analysis are worth further research.