I’ve seen several aficionados of the pure sciences dismiss more humanistic sciences such as economics or behavioural science. More often than not, they miss the point.
A map is a model of the territory. A map is wrong – it doesn’t capture every detail in the territory. Yet, despite their inadequacies, maps help us get around. Every theory we have is a model of how the world works. All those models are wrong – they break down at certain boundaries (even the theory of relativity and quantum mechanics). Yet, some models are useful.
Every map has both a region a shelf-life. A map of a city drafted 100 years ago is useless – the city has moved on. Maps that outline countries remain accurate longer. Longer still are the ones that outline geographical features such as lakes, rivers and mountains.
Most natural sciences are maps with large regions and shelf-lives. They describe systems that move at the slow speed of nature. Therefore, their models are more deterministic. Newton’s laws are universal – they are just as applicable here as they are in the outer edge of the universe, and they were just as applicable soon after the big bang. They are also deterministic – you cannot refute newtonian principles unless you really pick nits.
However, theories that describe human behaviour, like economics or psychology, describe smaller regions and have lower shelf-lives. They apply mostly in a given cultural context, which changes with each hundred kilometers. Their systems also change at the pace of culture – almost every generation or so. Therefore, their models are less deterministic and more probabilistic. Almost every theory in behavioural science, even the most famous ones, face a replication crisis.
These theories are often wrong, but they can also be useful. Behavioural science can give you guidelines on how to streamline road traffic, help people save more money and convince a state’s population to get their Covid-19 vaccine shots on time. Good luck trying to achieve those things with newtonian physics.
All models are wrong, but some models are useful – it is important to balance this wrongness with use. The higher its use, the more wrongness we must learn to tolerate.