Not perfect

This is perfect: walking with dog along city shore

Holiday baking can feel routine after many years of practice. And yet, yesterday I binned a batch of cookies from a recipe that I’ve been making for decades. I would have thought I’d have it down by now. Clean the pan. Move on.

Over a lifetime I’ve become more practiced in accepting my own failures. Decades of developing software to pass multiple levels of testing was an ongoing study in acknowledging failure. The testing of complex software inevitably discloses lapses in anticipating variability in data and user behavior, or inconsistency in logic, or simply design flaws. The same can be said of any of the creative arts, where critical feedback – objective or not – accompanies every presentation or performance. A client may reject a commissioned work after the artist has completed her work; the artist’s and client’s visions did not align.

Scientific research by definition invites imperfection. A study would be suspect if every test of an hypothesis validated the hypothesis 100%. There would be nothing to learn from such a study, and you might reasonably question whether the tests or data had been manipulated.

Motivational posters attest to this homely message:

I have tried 99 times and have failed, but on the 100th time came success.

Albert Einstein

The message would appear to be liberating: It’s okay to fail! You don’t have to be perfect! Despite the exhortations, failures large and small can be hard to bear. There is a cost to failure. There is loss. Objectively, it may just be the cost of materials, or the loss of the time investment. Yet it feels like more than that. It can feel personal as you accept that your notion of your identity or your cause has been eroded, even ever so slightly.

Canine scent detection for human health is proven and verifiable. Despite this, every scent-trained dog won’t be 100% accurate on every trial. Similarly, no human-devised health screening technology is 100% accurate, even though these are foundational to modern medical health maintenance. For instance, PSA tests have a specificity rating (accurate detection of cancer presence) of 21%, and yet this method is routinely used for prostate cancer screening. Our current concepts of modern medical health screening are such that weak accuracy percentages don’t challenge them.

This tolerance doesn’t necessarily extend to new models, which seemingly must be flawless to be considered as alternatives. Yet radically different technologies (such as canine olfactory and cognition processes) needn’t be perfect. Inconsistencies provide opportunities for learning how dogs can discern scents imperceptible to us, so that we can better understand limitations and refine training and testing protocols. The real question should not be whether it’s perfect, but whether we’re better off for it.

So far, the answer is perfectly clear.

Leave a Reply

Your email address will not be published. Required fields are marked *