There is no proof (except when the authors wants there to be)

by Carl V Phillips

I finally have a few free minutes while others tinker with the document I have been working on, so I thought I would comment on something that struck me about a post from Chris Snowdon this morning.

In the post, Snowdon rightly ridicules the UK National Health Service for bragging about their programs to assist smokers who want to quit, which he points out reached very few people, with minimal success, and at rather great expense.  I found myself taking the thoughts one step further, and questioning whether even the claimed modest NHS accomplishments really occurred.

What struck me is that the anti-tobacco cadre who claim that this (and other similar smoking cessation programs) are a success is the same crowd that will deny clear evidence of THR success, based on such claims as “correlation is not causation; we do not know anything until we have randomized trials”.  As in, there is no proof that Sweden’s low smoking rate is due to substitution of snus, or that reductions in cigarettes sales that match increases in e-cigarette sales mean that e-cigarettes are replacing smoking.

In this particular case, there is apparently not even a correlation.  The NHS provided a service, and a tiny number of people who availed themselves of it quit smoking.  Did the success rate go up with intensity of effort?  It does not sound like it.  Did those in the system quit with a greater success rate than similar people not in the system?  Perhaps, but quite probably not — but NHS does not seem to have even tried to figure that out before declaring victory.

Of course, these tactics — believing anything that tends to support one’s position and denying everything that does not using made-up-on-the-spot rules about what evidence “counts” — are common across all dishonest activists.  The same tactics are used by those who wish to deny the health risks from cigarettes.  Hardly a day passes when I do not see these tactics used by spokespeople for some industry (including, I should note, the e-cigarette industry though, to my recollection, never in modern times by the major tobacco companies).

Here is the flyover takeaway message from this:  There is no proof in empirical science (i.e., there is no proof in the real world, only in constructed systems like mathematics).  We sometimes use the word to mean “overwhelming evidence”, of course.  But anyone who tries to make an argument about what is true in the world that hinges on the concept of proof either does not understand what they are talking about or are trying to mislead you.

Almost any data allows us to make some inferences about cause and effect.  But no data ever proves it.  Moreover, there are no simple recipes — analysis is required.  Some data are easier to interpret better than others, but data is not the same as knowledge.

When someone claims causation merely because the data shows one event followed another (as with the NHS), that is clearly not adequate analysis.  But at the same time, when someone just cries “correlation is not causation!!!” with no further analysis to explain why the particular correlation is apparently not causal, that is arguably even worse.  Don’t trust anyone who makes a habit of doing either one of these.  Anyone who does has demonstrated to you that they are not seeking to discover the truth, let alone to tell you the truth.


8 responses to “There is no proof (except when the authors wants there to be)

  1. Carl, what do you think is their motive is for fighting against THR?

  2. Hard to trust anything the antis say these days.

  3. Yes, I agree: there is no proof in empirical science. But, we can “prove” some propositions. For example: Is smoking a necessary condition for lung cancer? We can prove that this proposition is false. Is smoking a sufficient condition for lung cancer? This proposition is also false. And it is because we know that these are false propositions that we do statistics. Do you agree Carl?

    • Carl V Phillips

      Well strictly speaking, no, but I do see your point. Even for necessary or sufficient, no amount of worldly research can rule out the possibility (to the level of true proof) that what we have observed can be explained by, say, measurement error (e.g., every observed case of lung cancer in a nonsmoker was an error, such that it was not really cancer or he actually smoked). That said, it is true that for a tightly drawn conditional (which is what “necessary” and “sufficient” are), as opposed to “cause” in general, that we come much closer to proof. (And I admit that “closer to proof” is a rather sketchy phrase.) Indeed, if you go a step further and phrase the measurement itself into the definition (e.g., “someone has AIDS if he meeting certain conditions of immune system failure and HIV is detected in his body“) then you have an even closer “almost proof” that HIV is a necessary cause for AIDS; it fails to be a mathematical (logical) based only on even more extreme possibilities, like “someone did indeed diagnose the HIV, but lied about it in the lab report”.

      Setting all that aside and returning to the practical, we can often rule out necessary or sufficient so compellingly with very little data, such that when we consider all the data we have, there is absolutely no room for doubt among non-crazy people. (Note: there are a lot of crazy people.) This is seldom true in epidemiology for cause in general, where some of the time the outcome occurs only because the cause was present. There are exceptions, of course. Car crashes cause trauma and cigarettes cause lung cancer — thee is no room for honest doubt, given all the evidence.

      If someone casually uses the word “proven” for things like that, it is not a problem. The problem comes when someone protests that compelling evidence is “not proof” which is always true, but conveys the (often false) message “there is room for non-crazy people to doubt this”. There is no room for non-crazy people to doubt that Swedes do not smoke much because they use snus. I would say that we are quite close to being able to say the same about e-cigarettes lowering smoking rates, though I will grant an exception for people who simply do not understand the data we have. Still, even for those who do not understand the evidence fully, the case is compelling, and so “there is no proof” is clearly a lie (a statement designed to make the reader believe something that the writer knows is not true, or is not expert enough to know whether it is true or not).

      Hey, I like that definition. I think I will add it to the tabs at that top of the blog. Amazing what a seriously sleep-deprived mind can come up with first thing in the morning.

      Oh, also I am defining “crazy person” to include anyone whose weight on their prior is 100% (to put it in Bayesian terms), also known as derp (to put it in the terms of the day).

  4. And I agree unreservedly with your first paragraph. It is for these reasons (among others) that I put the word “prove” in quotes. More: in science we need to agree about what we are measuring, and we also need to agree on what is a proof. Even in mathematics and logic we have different concepts of what constitutes a proof. Regarding the remaining paragraphs, I have some minor disagreements, but it’s a long story (regimes of truth, etc.). Great post Carl.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s