Delingpole: Why Must Coronavirus Decision Makers be Clouded in Secrecy?

Britain’s lockdown is based on the computer modelling of a scientist with a track record of failed predictions and whose widely-criticised methodology is subject to such secrecy, he still has not released the source code underpinning his model.

Professor Neil Ferguson, director of the MRC Centre for Global Infectious Disease Analysis at Imperial College, London, is the scientist whose doomsday modelling — predicting up to 550,000 excess deaths — frightened Prime Minister Boris Johnson into imposing sudden and radical lockdown measures across the United Kingdom.

Five weeks into the lockdown — and with still no indication from the government of when it might end — increasing scrutiny is being focused on the scientific credibility of both the doomsday Imperial College study and of its lead author, Prof Ferguson.

Among the questions being asked is why — despite having promised to do so on Twitter — Prof Ferguson has still not made public his source code.

On March 22nd, Ferguson tweeted:

Rather than release his source code immediately, however, Ferguson deferred the issue while blaming pressure of work…

…and illness.

But still, there was no sign of Ferguson’s source code — the source code, let us not forget, on which depends the British government’s decision to hold most of the population under virtual house arrest and to close down a significant chunk of the world’s fifth-largest economy.

While it is true that Ferguson has released some code on Github, it cannot possibly be his original source code because, as he stated in the tweet above, that code was written in C.

The code published on Github is written in another language (R; plus some Python), suggesting that it is not the source code but a reconstruction of the source code.

Furthermore, the code published on Github is clearly incomplete. In order for published codes to have any value to scientists who wish to reproduce those results, they need to time-stamped at every stage of development (known as a “commit”), so that each amendment can be noted and checked, and every error (if any) traced back to its original source.

The Imperial College doomsday paper was published on March 16th. But the earliest commit in the Github code is on or about March 27th. Clearly, Ferguson and his team cannot have begun writing software to model the Covid-19 threat after they published their paper. So where is the evidence of their workings?

This is not an esoteric issue. For any scientific paper to have validity, a number of fundamental principles must be observed — or the paper is next to worthless. First, the methodology must be sound. Second, it must be reproducible — that is, other scientists must be able to check its workings and confirm that it is valid.

As things stand, this kind of software is known in the programming community as a ‘black box programme’ — you put numbers in, and other numbers come out, but nobody can see what’s going on inside and how or why it works. This is an advantage in some applications where you want to protect your code from plagiarism, but when it comes to making policy decisions for nations, it means no accountability.

What is clear from the responses on Twitter that software programmers and engineers are highly unimpressed by Ferguson’s failure to observe at least the second of these basic protocols.

One software engineer I spoke to likens the situation to an architect building an extension to your house and then, when the extension falls down, being unable to provide the original blueprint he used to design the extension — but promising instead to provide you with a reconstruction of that original blueprint.

“If he won’t release the source code that means he cannot stand by his work — and that is unforgivable,” the engineer said to me. “The taxpayer will have paid for that model. It is simply the most basic professional ethic that in return for the public money spent on his work, Professor Ferguson should show how he reached his conclusions by publishing his code in open source form.”

Since so much depends on the validity of Ferguson’s modelling, it is understandable that there should be such concern about his methodology and track record.

Ferguson’s form, as LockdownSceptics notes, is not impressive:

• In 2005, Neil Ferguson told the Guardian that up to 200 million people could die from bird flu. “Around 40 million people died in 1918 Spanish flu outbreak,” he explained. “There are six times more people on the planet now so you could scale it up to around 200 million people probably.” The final death toll from avian flu strain A/H5N1 was 440. (That’s 440 people, not 440 million.)
• In 2002, the same Professor Ferguson predicted that mad cow disease could kill up to 50,000 people. Thankfully, it ended up killing less than 200.

But then, computer modelling of epidemics generally is notoriously inaccurate.

• The CDC’s model predicted that 1.4 million people would die from Ebola in Liberia and Sierra Leone five years ago. The final death toll was less than 8,000.
• The U.S. Public Health Service predicted that at least 450,000 Americans would be diagnosed with AIDS by the end of 1993. In fact, the number was 17,325.

Yet on Ferguson’s paper almost the entirety of Britain’s lockdown policy currently depends.