COVID-19: Mathematical Modelling on Trial

Mathematical modelling is at the centre of our lives as never before—invoked and presented daily to justify massive change in our way of life, livelihoods—even as giver of life or death.

It’s presented by experts not only as the best torch to pick in navigating us to the end of the COVID-19 tunnel, but the only one. It predicts the future or how we must change policy to achieve a different future.

This faith in mathematical modelling is a double-edged sword for maths’ future. Succeed in averting worse catastrophe and it will be hailed as our saviour, a force on the rise to yet greater heights, a direction to trust in more. Fail—driving us into health and economic deprivation—and its power will be eroded. Will we even be able to tell which it achieved given that every possible outcome will be bad at one level or another? With no control experiment world, even figuring this out will be hard…claims and counterclaims no doubt relying on mathematical modelling…or a stark pre-enlightenment, anti-science alternative.

You might expect my unequivocal support for maths in this crisis, a chance to prove its metal once and for all. Yet what we’re asking of its predictive power today is extreme: no less than telling us how two worldwide, multifaceted, intertwined complex systems (global health and economics) will play out. Worse, we are asking for individual answers, systemwide answers all with local details that almost certainly affect the large-scale outcome significantly. For example “going to work” might mean being on a crowded underground system in a major city (meeting many new people all the time) then working very close to people in an office all day or it might mean driving to work by yourself to work in a windy field with the same 2 colleagues every day—likely very different for how the virus spreads.

“Computational literacy at the very minimum offers some immunity against computational misinformation”

Maths isn’t the best game in town, it’s pretty much the only game in town. Its been thrust, even over-promoted, into a position that it didn’t or shouldn’t have claimed, depending on which expert is driving.

Some experts have deftly guided maths modelling’s role, ascribing it deserved power but not inappropriate omnipotence. Living in the UK, I have watched the chief medical officer Chris Whitty brief the public with clarity, honesty and real answers like “we don’t know” or “we can’t predict that” or we “need more data before we can say”—answers we may not like for their indefiniteness or lack of easy resolution but which garner respect. On the other hand some experts are offering up predicted (and usually shocking) numbers of deaths with a certainty that seems incongruous with their model’s sophistication: for example assumptions that COVID-19 works like flu, different locales work similarly or that viral load at infection makes no difference to a person’s outcome if they do get enough to be infected at all.

Take this last assumption. I’m by no means an expert, but I find it unconvincing. With unprotected ENT surgeons and ophthalmologists like my wife and others working very close to people’s faces apparently highly represented in “healthy people” deaths, is it really a correct assumption that a huge initial dose of virus gives you no worse a chance of a very bad infection than an initial dose just large enough to get you infected at all? If using binary assumptions “infected” or “not infected” in the models is wrong, the predictions will probably be badly wrong too. For example, it may not be so dangerous to have people slightly infected—even positive if it can build up immunity with low risk—but you definitely don’t want any possibility of a big group of infected people together. Or someone infected pushing virus down someone else repeatedly for a long time. But places where you’d only get mild exposure might be OK, at least for healthy people. I am not in any way saying I know if this is the case; I’m merely trying to point out how just one key assumption error (which seems plausible) could drown out otherwise careful maths. Worse if the assumption taken doesn’t easily sit with the public’s “common-sense” as you could argue my binary viral load bugbear one doesn’t, and it proves wrong, this it terrible marketing for belief in reasonable maths against the anti-science unenlightenment cohort.

There’s an even broader issue, of what we’re trying to achieve or optimise…meshed with what’s even possible? Is it containment, keeping under intensive care capacity in our hospitals, slowing the spread in the hope that treatment or vaccine will come along before more get infected, or natural herd immunity. Again, sometimes there’s been clarity and honesty, which the maths can hope to enumerate into actions to take—like reducing cases so as not to overwhelm UK NHS capacity—sometimes there isn’t.

These steps of sceptical and honest definition, abstraction, computation and interpretation are key to running the maths or computational process to inform better decisions, not overclaim its abilities. These are the very same steps that I believe it’s so vital we thoroughly educate our populations in understanding. My colleague Dan Robinson’s blogpost Tackling a Pandemic: A Computer-Based Maths Approach sketches out how they are being applied to COVID-19. They are the bedrock of acquiring computational literacy, as well as expert computational thinking—what I see as the AI age’s key building block for mass education, much as literacy has so successfully been over the last century. This education at the very minimum offers some immunity against computational misinformation.

These are key themes in my forthcoming book “The Math(s) Fix: Education Blueprint for the AI age” alongside proposals and plans for a fundamentally new mainstream maths education solution. It is only with this change that we can expect our populations to have judgements that allow them to better set or interpret rules, understand risks and for us all, collectively to avoid greater meltdown. No time like now to see its importance.

The jury will soon be out on COVID-19’s mathematical modelling. Before our populations retire to consider their verdicts, let’s put the case for computational enlightenment and its education not anti-science fundamentalism. Let’s use this shock to the system to reform the educational system, starting today.