Where’s the Black Squad when we need them?

Q: What do weapons of mass destruction have to do with cot death?

A: In both areas, the “experts” evaluating the evidence and acting on their conclusions have caused enormous devastation. Then, after the fact, that evaluation has proven wrong.

Why? I have an insight that may be useful.

I’ve held a number of jobs in my working life. The three that I’ve spent the longest at, though, are paralegal (2 years), financial auditor (3 years) and software tester (7 years and counting). Though they seem quite varied, they have one common factor: they’re all about the evaluation of evidence.

Lawyers and paralegals, of course, work with evidence all the time: gathering it, presenting it, writing about it. There’s no pretense of neutrality. A trial lawyer’s job (aided by paralegals) is to find evidence that supports one particular view, and to discount evidence that doesn’t.

Financial auditors are, on the surface of it, very different from lawyers. They go into companies at the year end and check the financial accounts those companies produce. Each stage of the audit is made up of tests on certain aspects of the accounts, whether it be a stock count to ensure that the inventory numbers are correct, or a check of reconciliation procedures to allow the auditors to rely on internal financial systems. And for each stage of the audit, we used to state the specific object of the test. I still remember the format.

Object of Test
To accumulate audit evidence that stock valuations are materially accurate and correctly stated in the year end accounts.

The public used to percieve auditors as unbiased and neutral (possibly even stringent and difficult to satisfy), but of course the scandals of recent years (the Maxwell empire, Baring’s, Enron) changed all that. Everyone knows the subtle, unstated pressure that the auditors are under when they go into a company, particularly one which pays the firm’s consultancy arm large fees. It’s almost unheard of for a Big Five firm to refuse to sign off a set of accounts.

Learning about software testing allowed me to see consciously what I knew unconsciously already. The audit process is biased in favour of approval, and any such bias makes an enormous difference to the results obtained. This is a phenomenon that testers are painfully aware of – it’s the reason that software has to be independently tested.

To quote one of the foundational books on software testing (The Art of Software Testing, by Glenford J Myers1):

“Since human beings tend to be highly goal-oriented, establishing the proper goal has an important psychological effect. If our goal is to demonstrate that a program has no errors, then we shall tend to select test data that have a low probability of finding errors. On the other hand, if our goal is to demonstrate that a program has errors, our test data will have a higher probability of finding errors.”

Reread the sample “object of test” above in the light of that quote. What is the goal of the test? Is it to find “bugs” in the accounts, or to establish that they aren’t there? How likely does that make it that we would find errors?

The most successful software testing teams are the ones who take a skeptical, or even hostile, attitude toward code quality. IBM’s infamous “Black Team” took this to extremes, dressing in sinister clothes, cheering when they found bugs, and deliberately striking fear into the hearts of the programmers whose code they tested. The reliability of mainframe operating code is their legacy – we wouldn’t have a hope of achieving “six nines” (99.9999%) availaibility had they not found the bugs they did.

So if bias affects results, how then do we view Professor Meadows and his eponymous Law (“One cot death is a tragedy, two is a coincidence, and three is murder unless proven otherwise.”)? His testimony has jailed women since acquitted of the deaths of their children, and caused authorities to remove babies from their parents, sometimes permanently. Yet statisticians claim that he took a “stamp collecting” attitude toward evidence, including the cases that supported his views and overlooking the others. And given the above axiom, how did he approach the deaths of children, when asked to testify at their mothers’ trials?

How a similar bias could affect the officials of two governments, when considering whether to send in the tanks, is left as an exercise for the student. But it begs the question: will any enquiry that focuses on the evidence, rather than the objectives of the people evaluating that evidence, really explain the conclusions that led to war?

  1. This is listed at $150.00 on Amazon at the moment. That’s pretty expensive, even for a computer book, but this one’s worth it. It’s 177 pages long and has been indispensible since its publication in 1979…quite a contrast to Martin’s Microsoft exam books, which can run to over a thousand pages and are obsolete before the ink dries.