Wouldn’t it be great if we could feed enough information about the world into a huge computer to predict what will happen 10, 50, or even 1000s of years from now?
This is what Hari Seldon, in Isaac Asimov’s novel Foundation does. Of course, as Asimov realised, we’d have to keep the predictions secret so people don’t interfere with the way things are supposed to turn out. Regardless of where you sit on the ‘It would be wonderful/awful’ spectrum, the very idea of predicting future states of the world continues to fascinate and perplex philosophers and social scientists. Why is it so difficult to make predictions about society? I suggest that the problem is not the complexity of the task, but the concepts we use to think about the world.
Wouldn’t it be great if we could feed enough information into a huge computer to predict what will happen 10, 50, or even 1000s of years from now?
The first obstacle is the one that Asimov tried to head off; that people reflect on the predictions made about them and alter their behaviour accordingly. For example, if we’re told that looking someone in the eye is a marker of trustworthiness, conmen everywhere will practice maintaining eye contact. Studies have also shown that studying economics changes how students behave in social science laboratories — they behave as the models they learn about suggest they should. This is not a problem natural scientists usually face. But, as Asimov suggested, in principle it is possible to keep predictions secret.
Other philosophers have suggested that this self-reflection is really just a type of complexity, and that complexity can be dealt with. For example, physicists would be hard pressed to predict the path a particular leaf on a particular tree will take when it falls to the ground. Physics usually works at a more general level. In the same way, so the argument goes, perhaps we are thinking about predictions in the wrong way; we might be able to make predictions about a large number of human beings (Asimov suggests c. 75 billion people), and we might make those predictions about neurological or other features which humans can’t interfere with.
Our ability to process large amounts of data has allowed us to discover some surprising regularities in human behaviour. For example, a formula has been discovered that predicts the number of pages an internet user visits within a website. Interestingly, this formula was developed in the 1990s and has remained constant (at least until 2006) despite the developments in the internet since then . The data from peoples’ mobile phones suggests that their location is highly predictable. Nevertheless, we often want to predict what people will do in a wider sense than this- will they rebel, or will an Empire fall?
Our ability to process large amounts of data has allowed us to discover some surprising regularities in human behaviour.
The second reason why it is difficult to find laws in the social sciences is familiar to every student of economics; ceteris paribus conditions. This phrase, seemingly inserted after every generalisation means ‘all other things being equal’. For example, ‘the price of apples will rise if the supply of apples falls’, is true as long as the demand for apples remains the same. To some extent, scientific laws also have ceteris paribus conditions attached; factors such as air resistance, temperature and friction often need to be controlled for during experiments.
However, in the natural sciences we know, for the most part, what the things are that must be held constant. In the social sciences it isn’t so clear. Political scientists generally agree that democracies do not go to war with one another (this is called the Democratic Peace Thesis) but they don’t agree about why this is; some attribute it to decentralised decision making, others to the economic structure of democracies. Without knowing why this is true, we can’t even begin to list the factors that need to remain the same. Ceteris paribus clauses in the social sciences are often seen as get out of jail free clauses that state that as long as anything that interferes with a generalisation doesn’t change, it will be true. This is in contrast with many, if not most, cases in the natural sciences where we know what is likely to interfere with generalisations and predictions. Not all philosophers agree about this though; some argue that the laws of physics are more like the laws of economics than we like to think.
These two problems are significant, but not necessarily insurmountable. Optimists amongst us might hope that increases in computing power might eventually overcome complexity worries. After all, it is possible that we could predict not only what people will do, but how they will react to knowing what other people think they will do. Ceteris paribus clauses might also become less troublesome as we get better at modelling human societies. Just because we don’t know why the Democratic Peace Thesis appears to be true doesn’t mean that we have to give up all hope of discovering this in the future.
The third reason why it is so difficult to predict human behaviour is that the concepts we use in the social sciences are odd and make predictions difficult. This is a little trickier to overcome than the first two worries. The paradigmatic concepts in the natural sciences, things such as ‘gold’ are easy to define; we can state what we’re talking about precisely in terms of atomic structure. This is in contrast with terms such as ‘poverty’ or ‘happiness’ the definitions of which, despite their prevalence throughout history, remain the subject of academic debate. This second sort of concept is often described as a ‘cluster concept’ — a concept groups together a cluster of features, such as lack of insufficient nutrition or income below a poverty line, but not all these features are necessary for being in ‘poverty’. The definition is likely to vary by context but despite this, different people can generally agree about when the term applies.
But, again, the contrast with the natural sciences is not as clear-cut as it at first appears. Concepts like ‘species’ in biology are contentious; there are a number of competing theories about how to define species, so much so that many biologists simply avoid using this term. Most usually, a species is defined as a group of organisms which breed to make fertile offspring. This is why horse and donkey are different species; they can breed, but the resulting offspring (mule) is infertile. This definition is problematic for asexual species though. Organisms can also be classified according to DNA, body types, or according to the ecological niche they inhabit. Darwin put it well when he said, “No one definition has satisfied all naturalists; yet every naturalist knows vaguely what he means when he speaks of species.”
Isaac Asimov: Foundation.
A multi-volume galactic empire saga. Many say, Asimov’s most important work.
Amazon affiliate link. If you buy through this link, Daily Philosophy will get a small commission at no cost to you. Thanks!
This is relevant to predictions because it is difficult to predict the future if we can’t agree on what the terms we are using mean. While Darwin is right that we know vaguely what we mean when we talk about poverty, or happiness, or species; this is a big problem once we try to make predictions.
Consider the prediction that poverty will be eliminated within 100 years. How do we judge whether this prediction came true if we don’t agree about what poverty is? Suppose that our world goes through a period of development, such that people live to some maximum life expectancy, with an adequate supply of resources, suitable housing, good education etc. Has poverty been eliminated? It depends on who you ask. Poverty is relative, and has many meanings; we have begun to talk of technological poverty in recent years, which would never have crossed people’s minds 100 years ago. However, in our present society, access to technology affects our ability to function. Those lacking access to the internet are increasingly deprived of banking and governmental services. This is arguably a sort of poverty. Who knows what other types of poverty will arise in the future?
Consider the prediction that poverty will be eliminated within 100 years. How do we judge whether this prediction came true if we don’t agree about what poverty is?
Hari Seldon made a different sort of prediction; that the galaxy would fragment, beginning at the outermost fringes, and that a particular region would split into four kingdoms. Seldon’s predictions seem capable of being disproved: we need four kingdoms, no more, no less. However, if he had been even slightly less specific we’d have a problem. How much fragmentation is enough? What does fragmentation look like anyway?
If we want to make predictions, or formulate laws about human behaviour, then we must use terms that are specific, and about which agreement is possible. There will, of course, be evidence, and things upon which future social commentators can agree. What is more difficult is deciding whether the evidence is sufficient to agree that, for example, a revolution occurred or a system fragmented. This is because the concepts ‘revolution’ and ‘fragmentation’ are so vague, and applicable to many different sorts of situations. For example, the Cambridge Dictionary definition of revolution is ‘A change in the way a country is governed, usually to a different political system and often using violence or war’. How much change is enough, and how different must a political system be? What counts as ‘violence’? There are also technological revolutions. Someone trying to make a case that a revolution occurred is quite likely to be able to do so. This means that, unfortunately, future social scientists would probably get bogged down in ‘yes there was/ no there wasn’t’ debates. Much as current social scientists debate how to measure democracy, how to define poverty, and whether the Bolshevik ascent in Russia was a revolution or a coup.
So, if you’re currently building a system to predict the future of humanity, stay away from the social science concepts we’re used to using and predict things that are easy to measure, like population numbers, or the percentage of all humans living on Mars. Unfortunately, those predictions don’t sound quite as exciting as the downfall of humanity… Alternatively, predictions can be so vague that they allow for easy confirmation. Nostradamus, a French astrologer living between 1503 and 1566, wrote a book in 1555 in which, some believe, he predicted everything from the French Revolution, Adolf Hitler, both World Wars, the Apollo moon landings and the September 11th attacks on the World Trade Centre. However, his predictions are so vague, and often poorly translated from old French, that a determined believer can convince themselves that he predicted any number of things. So, the other option is to make entirely vague predictions and hope that future humans will creatively fit them to the facts.
◊ ◊ ◊
Catherine Greene is a Research Associate at the Centre for Philosophy of Natural and Social Science at the London School of Economics. Her research interests are the philosophy of finance and social science. Before studying for a PhD she had a career in finance and still consults an ethics and investment strategy. More information is available at www.catherinegreene.co.uk
Cover image by Drew Beamer on Unsplash.