The Model

On April 23, I was in trouble. I thought I could make a common sense program that shows how good common sense ideas mimic reality. For example, how well such ideas about the COVID-19 epidemic fit the observed consequences. That program is now available. But the process that led to it made me think further. Why am I doing this? Do I secretly have Trumpian ambitions to produce a panacea? No. Do I sometimes think that, without being skilled in microbiology or higher mathematics, I can make better models than the specialists? Actually yes. Because that depends on the goal and expectations. And so I think I can do that from my personal perspective. Even worse, as a responsible citizen I have a duty to understand as well as possible what is going on with the COVID-19 crisis to determine my democratic (or anti-democratic) attitude, with the knowledge of but not directed by the specialists.

I make the program (and the models I release on it) during the wild ride that COVID-19 brings us. That’s an opportunity that allows for the kind of complex testing I already advocate, in the manner of a periodic weather forecast. We just need to clarify which powers are at play and what their values were yesterday.

On 23 March the Netherlands arrived at the stage where a serious form of lock-down has been issued. The one-and-a-half-meter society has been declared, theaters and public events have been canceled, the catering industry, schools and universities have been closed, contact professions are prohibited. Social distancing is the new standard.

It is now May 5th. After 43 days of social distancing the consequences are beginning to show: the pressure on intensive care is decreasing, as are the daily numbers of deaths and hospital admissions — a success based on a combination of knowledge and fear. But social distancing chills both economy and social life, and therefore also politics. Today there was a demonstration against the distancing measures that was ended by the mounted police.

This means that the initial model (algorithm 0) must be made sensitive to the forces acting for and against the measures. I give my plan on how I’m going to do that.

Figure 1: The naive model applied to the world and to the Netherlands

In Figure 1, two screenshots of Algorithm 0 are in action. On the left, the model has been applied to the world (with a population of approximately 7 billion) and on the right, the model has been applied to the Netherlands (with a population of approximately 17 million). It takes 34 one-week cycles to bring the model to over 7 billion. If the model were correct, everyone on Earth would be infected on August 14, 2020. This would already be the case on June 26 for the 17 million Dutch people. Again: today is May 5.

I notice a few things for now. The program is enriched with bells and whistles. It is now a platform for running different algorithms (e.g. Algorithm 0 or Algorithm 1) on different domains (like the world or the Netherlands) and under different parameter configurations. But about this later.

I expect that algorithm 0 will deviate from reality, but that it will fit reasonably in the beginning. What I want to do is calibrate the underlying model on the basis of available observations about the number of infections and the number of deaths, and adjust it based on the political measures taken and planned and on the economic and social impacts experienced and expected.