Of breathalyzers and false positives…

At first, I thought that there had been an accident.

As I inched my way up the line to the police officer with the flashlight, I realized that I was caught in a RIDE spot check – an Ontario sobriety testing program designed to catch drivers driving under the influence of alcohol. There were flashing lights and pylons everywhere and the cars ahead of me were being channeled into a single lane.

I thanked my stars!

I had just passed up on an offer for a quick beer from a colleague.I know that one beer is below the legal limit set by the province, but, who knows what breathalyzers can do? The last thing you want is to end up on the wrong side of a breathalyzer test.

Some recent reports in the media had called into question the integrity of the breath testing equipment – Intoxilyzer 8000C – used by the police in in Ontario. Aging equipment and lack of maintenance procedures had culminated in a few disputed charges.

Imagine a scenario where you are charged with a “refused to provide sample” because the breathalyzer did not register your effort to blow into it.

Apparently, it happens!

As much as I support the idea of keeping drunk drivers off the road, the thought of innocent people being wrongly judged by machines is not a comforting one.

But then, this post is not really about RIDE programs and its merits or demerits.

It’s about false positives and false negatives, in life.

Which is worse, a false positive or a false negative?

It depends.

In the context of our legal system, a false positive is worse than a false negative.  The presumption of innocence is a cherished characteristic of modern justice and best summarized by Sir William Blackstone’s saying:

“It is better that ten guilty persons escape than that one innocent suffer”

However, in the context of medicine and diagnosing illness, a false negative can have dire health consequences for an individual. At the same time, false positives can create panic and divert scarce resources towards wild goose chases.

In one form or the other, we have all experienced false positives.

Whether it is your car alarm going off for no obvious reason, or being uneasy about the big burly guy walking behind you in a dimly lit street, we hold statistically unsound presumptions that lead us to judge situations and people around us.

Think of how your internal cognitive detector kicks in when you deal with someone who you believe is not telling the truth. The goal of the tools, whether cognitive or electronic, is to enhance our predictive powers that help us get closer to the truth.

But, like the breathalyzer, our cognitive detector is also prone to errors of judgement.

In most cases, finding truth is an iterative process. You get closer to the truth every time you discard a known piece of “false.” Under a more philosophical lens, truth should be viewed as a destination, and the things that we identify as false, the vehicles to get there.

Back to the police breathalyzers.

One of the issues with accepting the results from this device is that it has no published “uncertainty of measurement”- or as we ordinary folks would call it, an “error rate.” So, if we assume that the device is 95% accurate, that would mean that 50 out of a 1000 people stopped could get charged incorrectly.

Additionally, if you assume that 1 in 1000 drivers are actually drunk while driving, the probability of an actual drunk driver being stopped and being charged is a mere 2% (1 out of 50). That would mean that the rest of the 49 could be incorrectly charged.

Tough luck?

Which brings me back to the top of the page, and the original scope of this piece.

The era of personal diagnostic tools is upon us and will soon dominate the consumer market in the same way smart phones currently dominate communications and transactions.

One technical obstacle remains: how do we accurately measure low frequency occurrences of a “something” within a population?

In the breathalyzer example above, a 99.99 accuracy rate would be needed to avoid falling into a false positive trap and mislabeling a driver “drunk” when they are not. We will need to have a breathalyzer tool with 99.99% accuracy.

Perhaps we do, but without a published error rate, we are in the dark.

At the end of the day, it is not about the breathalyzer either. It is about perceptions. What if someone was charged in error?

Are we too quick to judge?


Joe Totino

Joseph spends his time waiting for summer to arrive. He has interests in cognitive psychology and believes himself to be a decent chess player. He is based in Toronto and is an Engineer and management professional who works in the Telecom industry.

Leave a Reply