Inferring the infernal

I saw this photo on my Twitter feed. It made me laugh. Not because of the talkish gibbering on the sign, but because I read it as “back pain” and had to do a double-take.

I probably need to get my hearing checked out too. I often find myself not quite catching certain words - which results in me looking momentarily like an imbecile as my brain struggles to fill the gaps and match up what I thought I heard with what was probably actually said.

I wish I could remember more of my classic mis-hearings, but one I do remember occurred when my youngest daughter visited me for a couple of weeks in the country I was working in at the time. Driving back home after a day sightseeing I asked “what should we have for dinner tonight?”

Her reply, or at least the reply I heard, was “we could heat up the pagodas”

Of course, what she’d actually said was that we could heat up the leftovers. My brain completely failed to process this one and I had to ask her, after far too long a pause, what she’d actually said.

My mum who, God Bless her, struggles with poor hearing is even more legendary. Like the time she took her car to the local garage because it was making this horrible whining noise. The guys there spent ages trying to work it out. They couldn’t hear anything wrong, but mum was insistent. They eventually sent her home with a promise to look further if it continued. On the way back she adjusted her hearing aid and the engine whining miraculously stopped.

Fair play to mum, though, she turned round and went straight back to the garage to explain. Now when she takes her car to the garage they first ask, with huge grins, whether she’s checked her hearing aid.

You’ll notice that what I’m talking about here is one of inference. There are probably more fancy technical names but the essential problem is one of trying to work backwards to ‘reality’ from certain data - data that itself may not be all that perfect.

This kind of thing was described centuries ago by Plato in the allegory of the cave. Here, there’s a bunch of people looking at shadows on the cave wall believing this to be ‘reality’. The enlightened and elevated philosopher, of course, looks at the shadows and tries to figure out the reality of the shapes that are casting the shadows.

You can trust the experts to figure out what’s really going on with those pesky shadows.

There is a problem with this process though - especially when you’re working from limited, or partially corrupted data. The problem is one of uniqueness. There may be many possible 3D shapes in reality that will cast the 2D shadow on the wall, but which is it?

Inference problems abound in the scientific and technical world. X-ray crystallography, for example, really is a bona fide “shadows on the wall” problem. You fire a bunch of X-rays through some crystal and look at the ‘shadows’ cast. You end up with a diffraction pattern and the job is to figure out the internal structure of the crystal that caused that pattern.

In communications you send a message on some channel and the message is subject to errors (we might say that noise gets added). The job here is to figure out what was actually sent from the partially garbled message, the “shadow”, that was received. There are some really beautiful techniques for coding data that can achieve error-correction (at the expense of bandwidth) that rely on picking the most likely version of what was sent.

With the pantodemic we’ve seen the disastrous consequence of inference - or perhaps I should more accurately say we’ve seen the disastrous consequence of fixating on only one possible ‘reality’ from the many we could legitimately infer.

When we’re faced with several different hypothesised ‘realities’ that could explain the observed data (the problem of non-uniqueness) sometimes the best we can do is to form some judgement about which of those possible ‘realities’ is the most likely explanation.

Here’s an example of an inference drawn from the ‘shadow’, in this case the covid mortality data

I’ve just used a simple bell curve here to illustrate the death data and the kind of reasoning that gets used.

Proponents of lockdown might argue there is a very strong correlation between lockdown and the turning around of the mortality curve that can be seen in the worldwide data. They would be correct in that. They would argue that this demonstrates the efficacy of lockdown.

But we know that disease outbreaks are self-limiting. They have to be. In the extreme case where everyone eventually gets infected (and we assume, because of acquired immunity, there is little subsequent re-infection) things will come to a natural halt - there’s almost no-one left to infect.

This is an extreme case, though, and in reality our experience with previous outbreaks shows us that things turn around well before this extreme ‘everyone infected’ position. This is easy to understand. Initially, for a relatively novel virus, you’re going to see a rapid increase in infections because a lot of people are going to be somewhat immunologically naïve to this new(ish) virus. So things look like they’re getting out of control. As time goes on, however, someone with the virus will meet more and more people who now have immunity - the opportunities for transmission get fewer and fewer as time progresses. So things naturally tail off.

So, with this bell curve and the lockdown, are we seeing any lockdown effect at all? Maybe - but in order to prove that you need to see what happens when no lockdown is imposed - and compare things. That’s not going to be easy, either. As I briefly mentioned in my last post you’re going to have to be careful to compare “like with like” and account for any biases when comparing 2 different populations.

The first thing to note is that the correlation between lockdown and turn-around is built in anyway. Governments are more likely to decide to impose lockdown when things are looking bad, so lockdown (or at least the decision to lockdown) will likely precede the turn-around. If the curve dynamics are fundamentally driven by the process of VGV (virus gonna virus) then the correlation between lockdown and turn-around will be entirely spurious - simply an artefact of panic.

I don’t want to get into the nitty-gritty of whether lockdowns worked or not here. In my view they achieved very little other than harm. The evidence does not suggest a significant beneficial effect in terms of “saving lives” - quite the opposite in fact. What I mean when I say that is that the “shadow” cast by covid is better explained by VGV than by PGP (policy gonna policy) - it’s more likely the shadow cast comes from something that is predominately VGV shaped.

I am (clearly) not much of a philosopher. But, for me at any rate, I rather like the shadows on the wall analogy as a decent description of the scientific method. We see a shadow on the wall and build some construct, some mental model, of the ‘reality’ we think has cast the shadow. We test out our idea by seeing what shadows get cast when we change things up a little - do we see the shadows we expect to see if our ‘reality model’ is correct? If not, we have to admit our initial ‘reality model’ was not quite right - it did a good job of explaining one shadow, but didn’t work when we cast a different shadow.

Physics is very good at casting sharp “shadows”. In experiments we can often control everything and just let one variable change - this narrows down the possible ‘realities’ that are contenders. In other fields, like biology and medicine or the social sciences, the shadows cast can be more blurry and less sharp - and it’s more difficult to figure out what’s casting the shadow.

It’s critical to the process of ‘science’ that we examine the various possible “shapes of reality” that can cast a particular shadow. When we have a ‘shape’, like classical mechanics for example, that has explained so many shadows there is a resistance to some new ‘shape’, like quantum mechanics, that explains more shadows. There can be a resistance to an emergent better ‘narrative’ in science that can look a bit like suppression. Scientific ‘consensus’ is a reasonable rule-of-thumb metric at times, but it can be utterly, utterly, wrong - and we should never forget that.

What we’ve seen with the shadows cast by covid has been altogether different. Only one ‘shape’ has been allowed as contender for an explanation. Only one possible inference (or set of inferences) has been permitted. All others have been dismissed, or deliberately suppressed or ridiculed.

The cold, dread, ivory towers of academia turn out not to be so cold - there are passions and vested interests aplenty. But the arguments are largely confined to the weird and wonderful world of those afflicted with the personality disorder we label as “scientist”.

The covid battleground has been very different - it has occurred in the media, on social media, and by government diktat. It has also raged in academia too where many anti-narrative papers have had to face a significantly more stringent set of requirements to warrant publication whereas pro-narrative papers got rushed through (it was an emergency, see, so it was important to get this information out, but not that information).

But all this, too, casts a shadow. What are we to infer from the shadows of censorship and authoritarianism we have seen? What are we to infer from the widely publicized ‘mis-information’ we see (examples like 90% of those in ICU are unvaccinated, spread by UK politicians with absolutely no basis in the actual officially published data)? What can we infer from the egregious fear-porn used as propaganda by our governments?

Although several places are now rolling back the restrictions and seem to be going back to ‘normal’, I am not quite so sanguine. The easing of restrictions is certainly something to celebrate, but I think the underlying damage to our sense of balance and what is right has been severe. If I were to don the crinkly foil hat of cynicism I might even say governments are not in retreat, but actually very pleased with what they have achieved.

A very significant proportion of the population now think it’s entirely right and proper for governments to step in with sweeping restrictions of civil liberties, to determine who you can have in your own home, what you can do in your own home, how often and for how long you can leave your own home (and so on) - in order to keep other people safe from a virus that poses a low to moderate threat to the majority (and that’s overstating things). Many people think that health passports are a good idea - even though they might also recognize that, in the particular instance of covid, they haven’t worked too well.

Look at the number of people who have called for the severe restriction of liberties for those choosing to remain vaccine-free. Some have said they should be not allowed on the streets. Some have said they should be denied access to healthcare. Some have said they should be denied access to education. That this has become normalized and considered to be morally OK is not something to celebrate at all.

The shadows were cast, covid was a real thing, but the inferences that have been allowed to be drawn are truly infernal.

---