The Covid-19 pandemic demonstrated that model-based forecasts, which update only historical data, overestimate infection rates and weaken public acceptance of forecasts as a basis for policy decisions. Traditional models do not account for behavioural changes due to perceived risks and ignore the effects of crisis communication and subjective perceptions. The goal of SEMSAI is to explore how model-based forecasts can be adjusted to better reflect the future and how communication of forecasts influences behaviour. The DRU explores ways to increase the validity of simulation models for severe infectious diseases from a socio-psychological perspective. It offers evidence-based decision support for modelling to consider the public response and reflexivity in forecasting. Infection rates depend on the behaviour of the population. Also, published models and their interpretation affect the public response. Not considering these two aspects in simulations can lead to invalid models. Modelers need to understand people’s behaviour and feedback effects and find ways to integrate them (if possible) in simulation models. The subproject addresses these central research gaps by asking:

1. How do people behave during an epidemic?
2. What influences public response? 
3. How can public behaviour be integrated into modelling? 
4. Which feedback effects exist, and how relevant are they? 
5. How can we (if at all) integrate feedback effects into simulation models? 
6. Which conclusions can be drawn? 

Model-relevant indicators of human behaviour will be identified, integrated into models, and tested. The foundations for population behaviour and behavioural indicators during epidemics and pandemics are developed through extensive literature and document analysis as well as through secondary analysis of data from the DRU Covid-19 population surveys and complemented by a representative Germany-wide survey. The second phase of the project will focus on three survey experiments to analyse perceived model accuracy, communication of model uncertainties, and model content. Recommendations on whether and how feedback effects can be integrated into simulation models will be developed.