Computer-assisted models are getting better and better at producing simulations of complex future developments. Where do you see opportunities for data-driven forecasting?
The increasing availability and technical capability for data processing offer the opportunity to improve the modelling of social processes. The Covid-19 pandemic is a prime example. Modelling the occurrence of infections makes it possible, for example, to better coordinate vaccinations. When and how doctors’ offices can support the vaccination centers in vaccinating can be simulated on a daily basis.
At the same time, however, this also shows the challenge of every prognosis. The quality of any forecast always depends on the underlying data and assumptions. The less information is available and the more complex the underlying mechanisms are, the more difficult it is to produce reliable forecasts. Therefore, computer-based models quickly reach their limits when they are supposed to predict negotiation processes. If the decision options for actions are limited, for example in elections, they can be predicted with greater probability by a computer than, for example, the outcome of the monthly federal-state consultations on the pandemic.
Increasingly reliable forecasts provide decision-makers with ever more precise recommendations for action. How great do you consider the danger of an increasing lack of alternatives for decision-makers?
A lot depends on the available data. I still see a long way to go before algorithms pose the danger of a lack of alternatives. The contingency of the present is another factor. The mastery of politics is therefore to derive decisions from numerous forecasts, each associated with different assumptions and associated recommendations. The challenge for decision-makers is therefore not so much a “forecast monopoly”, but rather a profound selection between existing forecast options.
Political leadership balances the different scenarios. This process combines the accumulated knowledge of forecasting with the political experience of a decision-maker. Both together lead to decisions and their justification.
Pandemics, climate crises, criminology — algorithms-based models are increasingly used for forecasting. In your view, who should control the algorithms?
Algorithms, just like all other technologies, must comply with the legal framework. Political responsibility and digital competence from the public sector have the purpose of guaranteeing the fundamental rights of our democracy and the rule of law in the digital world as well. By having set up a data ethics commission, the federal government has taken this challenge seriously.
The government must now also empower the supervisory authorities to carry out controls in practice. What is needed are practical solutions that, on the one hand, give the supervisory authorities enforcement options, but on the other hand do not ignore the realities of the global digital economy (e. g. the protection of trade secrets).
Since the oracles of the ancient times, forecasting has been a tradition — in your view, how useful are forecasts in general, especially if they concern the more distant future?
The ancient oracles still balanced between fortune-telling and fatalism with a robust and humble view on the heavens of the Gods. It was not until the Renaissance, especially during the time of the great plague in the 14th century, that a deeper understanding developed to get to the bottom of the essence of chance. Chance now also reflects a chance of skill, of the controllable and the producible. With the Florentines Leonardo and Machiavelli, the intensive study of chance and thus of probability began. Leonardo summarized the basis of all future probability calculations in a famous saying: As you cannot do what you want, want what you can do.
Scientific forecasting as we understand it today only emerged in the second half of the 18th century in the context of the age of rationalism and mechanization. Since Machiavelli, strategic calculations have been indispensable for politics and an elementary part of strategy formation. Unlike purely scientific forecasts, however, they are not only based on universal and unchanging axioms, but primarily on empirical values and socio-political variables. A political system that never sleeps has enormous difficulties in making long-term political forecasts. There is a reason why political outlooks rarely go beyond the current election period.
An unshortened German version of this article was published on 12 March 2021 on the online debating platform “Meinungsbarometer.info”, accessible via this link.