How real are DNA test results?

Many services today use personal DNA to generate a wide variety of reports. But how real are reported results? And why do so many people see results that seem wrong? Recognizing that reported results are predictions helps answer these questions. Report services look at personal DNA and then predict something. All attempts to predict carry two risks: honest misprediction and pseudo prediction. We can judge how real report results are by the degree of both these risks.

Honest Misprediction

Most genetic reports use real science and are not completely made up. That scientific prediction makes some mispredictions should be no surprise. But these honest mispredictions can occur surprisingly often in genetic reports. Individual reported results need not be rarely incorrect to be scientific.

Take, for example, a car inspection service. It provides a personalized report based on your car instead of your DNA. A car inspection might report ‘at risk of brake failure’. There is a real scientific basis for this prediction based on the physical condition of the brakes. But the warning will still prove to be a misprediction for many unwise drivers who ignore the warning (the lucky ones).

Pseudo Prediction

Some reported results are too ambiguous and too amenable to wide interpretation. These predictions apply to almost all individuals no matter what their DNA. The results might be true, but are they really reporting something about your DNA if they apply to everyone’s DNA?

Imagine another car inspection service that reports your car ‘will get people moving’. The assertion "will get people moving" is easily interpreted as correct with ANY car. The inspection is a phony act of prediction. The report result is true, but it’s a pseudo prediction. Services can reduce the degree of honest misprediction by resorting to pseudo prediction. Reports will still feel accurate thanks to the Barnum effect (also known as the Forer effect). In another post to follow, you can see more examples of pseudo predictions and the Barnum effect.

Predictability Score

So, how real are reported results? That depends on the combined degree of both honest misprediction and pseudo prediction. Gene Heritage reports include a predictability score to help readers appreciate this combined degree. Three simple grades of ‘Highly Predictable’, ‘Fairly Predictable’, or ‘Barely Predictable’ summarize this score. Results at risk of being pseudo predictions only appear with a "Barely Predictable" score. A "Highly Predictable" score is only achieved with a small amount of honest misprediction. In a post to follow, see predictability scores for some familiar example traits. The next post will dive into the theory and mathematical details behind predictability scores.

Posted