The Italian Government decided to “manage” the Lockdown exit-strategy based on an “**Evaluation of reopening policies using social contacts and occupational exposure risk**“. This risk analysis, being based on a statistical model, is supposed to provide a “scientific” ground to support the operational choices to return to the (a)normality.

“It’s about time!” you could say, “finally someone decided to use the numbers correctly, instead of calculating percentages on unreliable data and deducing “trends” of mortality and lethality based on confused and unclear definitions!

The enthusiasm for a newfound rigour in policing-by-numbers, however, ends immediately after reading the first pages of the report:

The epidemic spread is simulated based on a stochastic SIR transmission model structured by age. The model takes into account the Italian demographic structure, the heterogeneity of social contacts at different ages and in different places of aggregation and the risk of exposure estimated for different professional categories.

What does that mean? It means that the scientific committee chose a statistical model whose results depend on the reliability of the information in input. If the data are good, the results are good too, otherwise, all you get are unreliable results.

This is made clear by **a study by Prof. Andrea Pugliese of the University of Trento** that explains how the spreading of infection can be assessed with the help of statistics. By reading this paper, it is clear that:

- to calculate the probability that a subject is still infected after the infection and that whoever comes into contact with him is, in turn, infected, it is necessary to know the time of the infection,
- the time interval between infection and onset of symptoms must be known, to calculate the incubation period’s length,
- the calculation of the so-called “serial interval”, the time distance between the onset of symptoms in an infected person and the onset of symptoms in an infected person, also depends on the possibility of knowing these values,
- the model works better if there is the possibility of identifying a time interval short enough to allow deaths for other causes to be ignored.

Another factor of intrinsic uncertainty in the results of an epidemiological model is that the evaluation of contagiousness can be estimated (I stress, **estimated**) in two ways: with the “**l****aw of mass action**” applied to the dynamics of contagion (anyone can infect anyone regardless of past contacts) or with **more complex models** that introduce more variables.

The law of mass action works for the molecules of gases which, even though “shaking” individually in a random way, on the whole, can be thought as all moving in the same direction. It is evident even to an inexperienced person like me that the application of this law to the dynamics of contagion is unrealistic. Still, citing the study of Prof. Pugliese textually,

the simplicity of the law of mass action makes it possible to carry out precise theoretical analysis and therefore to reach powerful conclusions.

Another methodological choice affecting the results produced by an epidemiological model is to decide whether the distribution of the contagion is “exponential” or not. I say “decide” not “calculate” because (I always quote Prof. Pugliese’s study)

The hypothesis that the period of infectivity follows the exponential distribution is tantamount to supposing that the probability of healing does not depend on how long an individual has been infected; as is intuitive, the actual distributions are usually very far from the exponential. But the exponential distribution has the great advantage of transforming the system (1) consisting of a differential equation and an integral coupled equation into a system of ordinary differential equations that are much easier to analyse and approximate on the computer.

In other words: we must choose to use a less precise but more “powerful” method or a more precise but less effective one.

Mind, I do not mean that statistical models are useless or that they shout random number but that, like all scientific theories, they have a value according to the hypotheses and assumptions based on which they are developed. They are useful and essential, but you need to know how to “handle” them.

Consequently, the fundamental thing to understand, when reading statistical models, is that they provide the tools to obtain values, but do not provide “truth”. In other words, once the model has been built according to certain assumptions and a correctly applied mathematics, the results depend not only on the “design choices” but also on the inputs.

Virtual the inputs, virtual the results all the same or, in a shorter sentence: garbage in, garbage out.

All that said, I would have expected the Scientific-Technical Committee to disclose the rationale of its choices, but it did not, as it just points out the assumptions with no further explanation:

In the model, we have therefore

assumedthat in the lockdown phase in Italy the social contacts outside home/school/work (i.e. in transport, leisure and other social activities)have been reduced to 10%of those observed in the absence of epidemic. The data provided by INAIL also suggest that 15% of workers use public transport. On the basis of this evidence, it is reasonable to think that in case of reopening of some productive sectors there could be an increase in the use of transport both by workers and by the public and we haveassumedthatcontactsdue to public transportwill increase to 20%. We have alsoassumedthat in correspondence with the opening of the commercial sector and the consequent necessary reduction of restrictions on the movement of people,contactsdue to “Other activities? (i.e. those due to frequenting shops and services)return to normal values (100%). We have alsoassumedthat leisurecontactsincreaseonly in the event of thereopening of accommodation and catering; on the basis of time use data provided by ISTAT, we havecalculatedthat time spent in catering establishments account for 24% of activities. We have thereforeassumedanincreasein this type of contactfrom 10% to 34%. It is assumed that there is no increase in contact due to outdoor activities, sports and recreational activities. (all emphasis added).

The document continues with a list of further “assumptions” (i.e. assumptions about the value of certain variables and not an objective measure of them) that populate the different “scenarios”.

Now, even “assuming” that the statistical model used is correct, if the initial values are “assumptions” (i.e. “arbitrary”) the results contained in the document have a purely virtual value. As William Thompson (better known as Lord Kelvin) said:

When you can measure what you are speaking about, and express it in numbers, you know something about it, when you cannot express it in numbers, your knowledge is of a meager and unsatisfactory kind; it may be the beginning of knowledge, but you have scarely, in your thoughts advanced to the stage of science.

But, as it was easy to predict (even without the use of “stochastic models”), the media have absolutised the largely “virtual” results of this report by spreading fear and uncertainty by stressing the upcoming possible increase of new overloads of intensive cares and casualty numbers rise at First World War levels. They did so, facilitated in their “mission” by the ambiguity of the official communications. The Government and of the Scientific Technical Committee should have clearly stated what the limits of this study were and why certain decisions were “taken”, instead of shielding themselves behind “science told me so” (at least because science did not actually do.)

One consideration emerges clearly from this story, and that is that complicating things breaks down the ability to understand them in the face of the progressive **“infantilization” of Western culture induced by the use of smartphones and tablets**.

Frankly, I don’t believe that this document was deliberately written and “communicated” in bad faith. It could, “simply” be a tool to formally sustain a political choice or to deflect its negative consequences and controversies by “**passing the responsibility of the decision on to science**“, as happened, for example, in the ’80, when t**he Ministry of Health declared the “legal potabilization” of water even though contaminated by high levels of atrazine**.

In general terms, it is certainly time to add to the first law of the apocryphal Goebbels, “repeat a lie a hundred, a thousand, a million times and it will become truth”, the second law of propaganda: “complicate enough a lie, and everyone will take it for true”.