[LINK] The government's coronavirus modelling

[Roger Clarke]

 > On Tue, 2020-04-07 at 18:10 +1000, Bernard Robertson-Dunn wrote:
 >> I can't believe they have been making decisions on the modelling they
 >> released today. They may be cunning but they are not stupid.
 >> IMHO, the big questions are "What is ScoMo hiding and why?"

On 7/4/20 6:59 pm, Karl Auer wrote:
 > ... I think they are genuinely concerned that if they told people how 
big this thing is and how bad it could get, they would have a panic on 
their hands.

Other people's guesses might be that:

(1) there are multiple models, and they paint quite different pictures

(2) the model(s) deliver very different pictures, depending on the
     nature of the data they're fed with.  That can reflect inadequate
     models (e.g. missing variables), or inadequate data (e.g. things
     not measured that - with 20-20 hindsight - needed to be measured,
     data defined one way but data collected in another, data definitions
     that changed over time resulting in incompatibilities, data that
     no-one ever thought to define that means whatever the collecter
     decided on an ad hoc basis it presumably was intended to mean,
     data that was of low quality when it was collected, ...)

(3) the model(s) deliver pictures that aren't consistent with what's
     been portrayed to the public

If the problem is (1) multiple models, that's normal science, especially 
in a complex, even wicked problem-space.  Try telling the punters that.

If the problem is (2) dubious-quality and/or inconsistent data, it's a 
question of sensitivity of dependent variables to differences in the 
independent variables.  London to a brick, this one's in play.

If the problem is (3) inconsistencies, it could indeed be that 'humanity 
was facing extinction'.  Or 'it wasn't so bad as to justify the measures 
that were taken'.  Given that it's been portrayed as 'pretty grim', the 
second of those alternatives is entirely tenable.

I don't want to sound unduly critical of the modellers, the data 
collectors, the data analysts, the public health tsars, or even the 
politicians.

This is all really, really challenging.

People confuse people-made models with real-world systems, and are 
encouraged to do so.  (Run a critical eye over weather forecasts).

A model is a simplified representation of some part of the real world.

When we find models that are pretty reliable, we justifiably celebrate.

But it's far, far rarer than the marketers (of all descriptions) want 
people to believe.

</sermon>


On 7/4/20 6:59 pm, Karl Auer wrote:
> 
> Two things. First off, they ARE stupid. They have demonstrated the
> depths of their stupidity over and over and over again (CensusFail,
> water rights, climate change denial, electricity privatisation, phone
> privatisation, NBN, RoboDebt - the list is practically endless.
> 
> In this case, I don't think malice is behind it. I think they are
> genuinely concerned that if they told people how big this thing is and
> how bad it could get, they would have a panic on their hands. They
> don't know how to release it and are terrified that if they do it will
> show they should have done more, better, faster.
> 
> It's paternalistic and, of course, stupid. It's stupid because the
> modelling WILL be leaked.
> 
> Regards, K.


-- 
Roger Clarke                            mailto:Roger.Clarke at xamax.com.au
T: +61 2 6288 6916   http://www.xamax.com.au  http://www.rogerclarke.com

Xamax Consultancy Pty Ltd      78 Sidaway St, Chapman ACT 2611 AUSTRALIA 

Visiting Professor in the Faculty of Law            University of N.S.W.
Visiting Professor in Computer Science    Australian National University