I attended a talk at a technical conference recently. The talk was intended to promote workplace diversity but bad mathematics from the speaker made the talk backfire and attendees came away with a negative view of diversity rather than a positive view.

Briefly, the speaker works for a large bank and pulled up the tired old statistically-infamous report that stated, “Companies in the top quartile for gender or racial and ethnic diversity are more likely to have financial returns above their national industry medians.” OK, fine, no problem. Except for neglecting to define how diversity is calculated. And neglecting to mention that the data includes only very large companies. And neglecting a dozen other things that make the statistics meaningless.

The speaker then read the next sentence from the report: “While correlation does not equal causation (greater gender and ethnic diversity in corporate leadership doesn’t automatically translate into more profit), the correlation does indicate that when companies commit themselves to diverse leadership, they are more successful.”

Wait a minute. What? To paraphrase the above: “Correlation isn’t causation but in this case correlation is causation.” Hmmm.

A few minutes later, an attendee asked the speaker something close to, “I don’t understand. Please explain how the data indicates diverse leadership makes a company more successful. Why not a much more likely explanation that successful companies have more money and can afford programs to increase diversity?”

The speaker had no answer and she basically went blank. It was very embarrassing.

I squirmed a bit when a second attendee asked, “If there’s a causal relation between diversity and financial returns, then the more diverse a company is, the more profitable it will be, but data shows this is not true. Can you explain?” The speaker had no answer.

This is an example of confirmation bias where people tend to see what they want to see and ignore any data that suggests a different point of view. For diversity in the workplace, companies could say something like, “We support diversity because we believe it will be good for our public image. We will give advantages in hiring and promotion to selected groups to increase our diversity.” But companies shouldn’t insult the intelligence of their customers by using bad mathematics. Or, opponents of diversity can simply say they think it’s morally corrupt to give advantages to any group at the expense of others, rather than pull up their own inconclusive data.

Confirmation bias occurs in many places. For example, experiments show that during the course of a trial by jury, jury members with opposite opinions tend to strengthen their beliefs as more information is presented, rather than use the information to come to a more rational conclusion.

There are all kinds of cognitive biases. The research I’ve read suggests that these biases evolved in humans over many thousands of years to deal with complexity.

Cognitive biases are one reason I’m usually skeptical about applying machine learning to any type of prediction or decision that involves human beings. Predicting which team will win a football game is one thing but predicting whether or not a person with a particular age, race, and arrest history will reoffend is quite another.

Ultimately, life is pretty simple. Doing the right thing is always the right thing to do. No bad statistics needed.

Reveal: I deliberately wrote this blog post in an ambiguous way, so that, depending on your preconceived ideas of diversity, you could interpret the post in multiple ways. Did you notice this effect?

Sometimes diversity is unambiguously good. This photo is from the infamous 2015 Miss Universe pageant when the host announced the wrong winner. Ouch.