Inspiration
Our country, Italy, was overwhelmed by the spread of COVID-19. As a group of students, we were unsure of what we could do to help out. With the help of Fabio Dercole, our Control Engineering Professor, we came up with this idea. Several teams of scientists around the world are dashing to model the spread of the novel coronavirus, at both local and global scale, with the common aim of hind casting and forecasting the COVID-19 pandemic. Standard epidemiological models are spatially extended to include multilayer contact networks describing communities, workplaces, schools, social and cultural places, healthcare facilities, and mobility within and across these layers. One important aspect is however systematically disregarded: our human nature. We believe that understanding human behaviour in a pandemic scenario is crucial for governments when drawing up public policy, as many of the related issues concern the encouragement and control of desired behaviours, lack of cooperation, and punishment.
What it does
The aim of Pan-EHub is to integrate game-theoretic behavioural and control sub-models into the available COVID-19 epidemiological models, and to test for different behavioural hypothesis and control strategies by means of both calibrated simulations and online virtual experiments. Experiments will be conducted by developing a mobile app and exploiting the principles of gamification. Participants will be immersed in a ludic multi-agent experience and faced, day-by-day, with simple information about the local and global social, sanitary, and normative situation and faced with simple decisions to make. Fully experimental results, calibrated model simulations, or data from a mix of human subjects and bots will enable our policy makers to forecast, among others, the effects of social lines of intervention, such as the level of incentives, control, and punishment.
How we are building it
We started by drafting an agent-based model that takes into account human choices and decision-making by providing the user with a daily list of actions to choose from with a point-reward system. This model is now being coded into python. The agents can either be controlled by people or bots on the server-side that make decisions using models from game theory. We are developing the front-end using Unity.
Challenges we ran into
The main issue for us was learning how to set up the server and the more technical aspects of the project, as we have limited knowledge in the field of computer science.
Accomplishments that we're proud of
We are proud of ourselves for having learnt to pick up relatively new programming languages, and for having tackled this project that was initially seemed too daunting too complex for our capabilities.
What we learnt
Simplicity is key, and that's especially true when scaling up an individual-based model on a large scale. You need to include the 20% of the features that will make the 80% of the difference. After quite some struggle and several sleepless nights, we got to a model that does surprisingly well with a fraction of the complexity.
What's next for Pan-EHuB
We look forward to widen our collaboration within our University first, and potentially the whole world after. We can't wait to develop and test our first version of the app, and then refine it based on the advantages that adding an actual human component to a computer simulation provides.
Log in or sign up for Devpost to join the conversation.