Almost all systems of justice and law recognize the efficacy of individual choice, and this is a good thing. At the end of the day, people who kill, assault, steal and defraud have the ability to choose their actions, and we should treat people according to the choices they make, for our own protection if nothing else.
Individuals make their choices in particular contexts, however, and anyone who believes that all human actions are equally likely regardless of context hasn’t been paying attention. Humans are social primates, and we respond to a variety of social and economic incentives in fairly predictable ways.
Consider a loaded die. One of the more clever kinds of loaded die has a cross-shaped hollow space in the interior, like this: +. It may have four arms or six, one extending toward each face of the die. A drop of mercury–a heavy metal that is liquid at room temperature–is enclosed in the hollow, and it can be coerced into any of the arms by tapping that side of the die down on the table. With the mercury drop in that arm, the odds are the die will tend to land with the opposite side showing.
There is no way to tell from one single toss if a die is loaded or not. Nor from two or three. For a six-sided die the odds of any one side showing up on a given toss are 1/6, or about 17%, so getting three of the same value in a row has a probability of (1/6)**3 or a little less than one half of one percent. So if the die was perfectly loaded it would take more than three tosses before you could be reasonably sure of it. Without looking at the distribution, there is simply no way to tell.
In reality, no die is perfectly loaded. It just has a tendency to land one way more frequently than fairness would predict. As such, it can take tens or hundreds of tosses before the bias is apparent. The die is still “random” in that you can’t predict the outcome of a toss, but it is what we call “weighted randomness”: some outcomes are more likely than others. The shape of the outcome curve is the probability distribution function (PDF) for the die. For a fair die the PDF is flat: all outcomes are equally likely. For a loaded die, some outcomes are more probable than others so the PDF has one or more peaks and valleys.
Physicists and other scientists deal with PDFs all the time. Whenever we measure anything we are dealing with the probability distribution of the results, and we have to worry about the shape of that distribution and its width.
Non-scientists often talk as if scientists measured “things”. We don’t. We measure probability distributions. If we’re lucky the distribution is narrow enough that we can ignore it, because it doesn’t really matter if the “true” value is 1.0000 or 1.00001 or 0.99999. In fact, it doesn’t even matter if there is a “true” value at all.
There is a furious debate amongst philosophers as to whether individual events within a random distribution have to be the result of some kind of mechanical, ultimately predictable, chain of causation, but this turns out not to matter very much. After all, the reason why there is still so much bickering about the question after 2500 years is that no one can tell the difference between a universe where all causation is purely mechanical and one where all causation is basically statistical.
In many practical cases the PDFs are so narrow that they can be safely ignored. My cat has a mass of 10 kg (really) and it doesn’t make a bit of difference if this is merely the peak of a PDF that might range between 9.9 and 10.1 kg. Whatever it is, it’s pretty narrow.
But lots of practical cases involve PDFs that are really broad, and that is when a failure to understand that there is a probability distribution behind every measurement we make can cause problems.
In the realm of social policy in particular, we have no direct control over anyone’s actions. That is, passing a law to make a particular act illegal does not mechanically cause anyone to do what the law says, and attempts to create such mechanical causation by way of the Total State have uniformly failed, in no small part because the cost of enforcement goes up exponentially while any improvement in compliance is incremental: if it costs $1 million to get 50% compliance, $2 million to get 75%, $4 million to get 87.5% and so on, the social cost of 100% compliance will be infinite, and something like this is what is actually observed in the real world, so the law of diminishing returns kicks in pretty hard when it comes to law enforcement.
Beyond that, laws that attempt to mechanically control the behaviour of others are based on the false notion that a bunch of middle-aged, middle-class white men in Ottawa have a better idea of what is best for me than I do, whereas I’m pretty sure I know a lot more about my personal situation than they do, and I trust my own judgement a lot more than I do theirs. The mechanical model of the social causation is based on the idea that the brain of Prime Minister Harper should be allowed to do the thinking for all of us on certain issues, and while I’m sure he’s a bright guy and all, his brain isn’t really up to that task: it doesn’t have the information it needs.
Furthermore, slavish compliance to the law doesn’t work: when Google first started experimenting with self-driving cars they programed to follow the law to the letter, and this proved to be a danger to everyone on the road. You can’t actually drive safely while obeying all traffic laws precisely. Remember that the next time you insist on perfect legal compliance as a goal to be aspired to in a free and just society. Like everything else, following the law has a probability distribution associated with it.
Given we have no mechanical control of other people’s actions and never will, a better approach to social policy is a probabilistic one. I don’t know why any particular individual might decide to do drugs or have an abortion or shoot someone, but I do know lots and lots of things that will change the shape of the probability distributions associated with those behaviours.
Medicalization and harm reduction approaches to drugs reduce usage and most importantly addiction. The War on Drugs does neither.
Easily available contraception reduces the rate of abortions, not abstinence-only policies and criminalization. Criminalization of abortion does not reduce the abortion rate: it only makes abortion unsafe for the mother.
Reducing the prevalence of guns and poverty reduce the murder rate. Criminalization also has a strong effect, particularly because we are quite properly willing to invest large amounts of effort into solving the crime, unless the victim happens to be an indigenous woman in the Lower East Side of Vancouver.
We, as individuals who have a responsibility for our individual actions, should advocate for legal and social policies that shift the probability distributions we care about in the directions we want to see them go. If we attempt to bring a mechanical-causation model to every single area of social or legal policy that we care about we are committing to huge expenditures for imperfect enforcement that frequently results in little or no effect on the problem we care about–this is particularly evident in the case of abortion–and will also do significant damage to a free and open society, as the War on Terror has in the US and the criminalization of drugs has pretty much everywhere it has been tried.
This approach does mean that there are people who will “get away with it”, whatever “it” might be. But 100% compliance with the law is almost never a good thing, because we live in a world of probability distributions, where the optimum has some breadth and uncertainty about it. Attempts to define the ideal more precisely than this will run up against the social and political equivalent of the Heisenberg limit, and the results of attempting to push beyond that are never pretty.