Using Game Theory to assess the effects of social norms and social networks on adolescent smoking in schools: a proof of concept study

In order to develop better public health interventions, it is important to understand the mechanisms by which they exert their effects. Interventions that act on groups of people or whole populations may be more effective at reducing health inequalities than those that target individuals. Changing social norms is one way of affecting behaviour in groups of people, but these social norms often depend on connections and shared "understandings" between members of the population. The connections between people are what characterise their shared "social network", and this may affect the way that social norms spread among people. Public health scientists know surprisingly little about how best to measure and evaluate the spread of social norms and their effects on behaviour.

This proof of concept proposal is innovative because we will harness the perspectives from scientists from different disciplines, for example, from economics and computer science, to compare and contrast two school-based interventions that aim to prevent smoking uptake among adolescents (13/14 year olds) in the UK and in Colombia - a LMIC and high income setting where smoking rates and norms are different - in order to reveal potential differences in the social norms-based mechanisms of action.

Overall Aims:
1. To improve the measurement of social norms around smoking behaviours in adolescents.
2. To use these improved measures to better understand the spread of social norms in school settings.
3. To better characterise the potential mechanisms of action of smoking prevention interventions in schools.
4. To learn lessons for the design and evaluation of behaviour change interventions that invoke mechanisms which change social norms.
5. To build a legacy of transdisciplinary research capacity in public health science in a LMIC setting, with clear pathways to impact.

Of the two evidence-based interventions, one is designed to harness the influence that peers have on their friends while the other is based on classroom teaching. In a before and after design, we will obtain baseline information on psychosocial characteristics, friendship groups and behaviours (e.g including attitudes and intentions towards smoking and vaping) from 300 students from three schools for each intervention (i.e. six schools in total, n=600) in the UK and the same number from six schools (three receiving one intervention each) in Bogota, Colombia. We are studying schools in two countries because the social norms around smoking are so different and the contrast between the two will provide richer explanations.

Before the intervention each participant will take part in economic experiments that will allow us to assess their judgments about the social appropriateness of a range of smoking and vaping-related behaviours, also enabling the estimation of individual sensitivity to social norms. At the end of the semester (after the interventions), participants will again take part in the same behavioural economic experiments, allowing us to estimate how the social norms (related to smoking and vaping) have changed, how individual sensitivity to classroom social norms has changed and how closely individual changes are related to changes among friends in the student's social network.

We will then analyse how measuring social norms in these novel ways has helped us learn more about the mechanisms of action of the interventions, and how the social norms spread in classrooms and year groups. We will use the quantitative data from the experiments and qualitative data from interviews with participants in these schools to "triangulate" our findings and sharpen the explanations that are possible from the overall analysis. At the end of the study we will meet with experts from different fields of research, to help us interpret our findings in a deeper way and discover how they might apply to other contexts and to other interventions to change health behaviours.

This proof of concept study will harness novel transdisciplinary insights to contrast two school-based smoking prevention interventions among adolescents in the UK and Colombia, where smoking rates and norms are different, in order to better understand social norms-based mechanisms of action. We aim: to improve the measurement of social norms of smoking behaviours in adolescents and how they spread in schools; to better characterise the mechanisms of action of smoking prevention interventions in schools, learning lessons for future intervention research. One intervention is designed to harness peer influence, the other is based on classroom pedagogy. In a before and after design, we will obtain psychosocial, friendship and behavioural data (e.g. attitudes and intentions towards smoking and vaping) from 300 students from 3 schools for each intervention in the UK and the same number in Colombia. Pre-intervention, participants will take part in a Rule Following Game, and in Coordination and Double Dictator Games that allow us to assess their judgments about the social appropriateness of a range of smoking-related behaviours, and the estimation of individual sensitivity to social norms. After the interventions, these behavioural economic experiments will be repeated, so we can estimate how social norms (related to smoking) have changed, how sensitivity to classroom and year group norms have changed and how individual changes are related to changes among friends. We will use Game Theory approaches to extract norms and norms sensitivity parameters from the experiments, examining the influence of individual student attributes and their social networks within a Markov Chain Monte Carlo framework. Putative mechanisms will be inferred by triangulating our experiments with qualitative data from participants, by having data that contrasts the effects of two interventions with putatively different mechanisms, and by the contrast between two countries where norms are different.

Our partners have identified a range of beneficial impacts, both through our regular meetings in developing the proposal and in their letters of support. We have designed our communication strategy (see Communication Plan) to maximise these benefits both for our partners and for our wider stakeholder community. Benefits range from improved understanding of underlying mechanisms of population level behaviour change interventions, through improving the approaches to their evaluation and informing future policy agendas. This range of benefits is briefly discussed below.

Promoting transdisciplinarity and ensuring stakeholder engagement will help us to deliver high impact research, with multiple beneficiaries in the UK, Colombia and beyond, including: young people; the general public; public health practitioners; policymakers across the UK and Latin America; participants of our studies and their families. Academic beneficiaries are detailed elsewhere (see Academic Beneficiaries). Considerable time and resources have been included in the grant to ensure that young people are involved throughout the study in order to maximise benefits to this population. Long term, we envisage such benefits to include the development and maintenance of 'healthy' social norms, leading to better quality of life and life expectancy through reduced smoking rates. Our Youth Advisory Panels will ensure that study findings are disseminated in a meaningful and appropriate manner.

The general public will benefit through increased knowledge of ongoing research. We aim to increase public engagement through events undertaken by our partners, Cancer Focus NI, the Public Health Agency, DECIPHer IMPACT and Foundacion Anaas. This type of engagement can occur throughout the research cycle, however, the public will also benefit, albeit at a later stage, from any research that is successfully translated into policy. Public health practitioners will benefit from knowledge gained through the workshop hosted in partnership with the Public Health Agency and Cancer Focus NI, and delivered as part of the implementation of the Public Health Agency's Knowledge Management Strategy and the work of the NI Public Health Research Forum (of which Kee was the inaugural Director). Parallel workshops will be planned for practitioners in the rest of the UK through our collaboration with the UKCRC Centre for Tobacco and Alcohol Research (www.ukctas.ac.uk/), Public Health England and the NIHR's Public Health Policy Advisory Board. These workshops will particularly focus on how best we can incorporate and elicit mechanisms of social norms in smoking prevention and other behaviour change programmes.

Policymakers and funders of research will find the research of value in assessing the public health priorities and novel approaches to smoking prevention. Although this is a proof of concept study, we have engaged with tobacco control experts from the outset, to ensure that this work will have long term policy relevance and impact. While the timelines for research and policy may not always be aligned, we will maintain our strong relationships with key decision makers to ensure that they have access to updates on new research and emerging results.

All beneficiaries will come together in a plenary workshop of thought leaders (from public health, network science and behavioural economics) using a Group Model Building (GMB), participatory approach that is widely used to build the capacity of practitioners to uncover hidden assumptions, build consensus and think more clearly about how changes might be effected within complex systems.