Fixing Bias

How do you fix bias?

More specifically, how do you fix bias in the systems we interact with every day?  That’s the theme of The End of Bias, an interesting new book by Jessica Nordell. She doesn’t just document all the conscious and unconscious biases we all have, but sets out to look at what works and doesn’t work in trying to address those issues. Well worth reading.

Plus, what’s not to like about a book that starts out with the experiences of a transman as he crosses the gender divide?

I wrote about this book a little while back – even before I read it! – and the simulation described in it; it shows how, even without overt acts of discrimination, a small level of systemic bias will accumulate over time and significantly disadvantage one group or another. 

And that’s a key point:  That you don’t need bad actors, overt discrimination or blatant wrongdoing to suffer from bias; it’s the small things that add up over time.

And key point two: We’re all biased.  None of us can escape the blind spots we have; and we all have plenty.  “Trying harder” isn’t a solution, any more than telling a nearsighted person to “try harder” to read the words across the wall.

So if more effort, better intentions, training people to recognize bias, aren’t the solutions – or at least not the only solutions – then what is?  That’s what the book’s about.  Bear with me. This is a long-ish post.

And all this matters not just as we try to build newsrooms that are more inclusive and more representative of the communities we cover and serve, but also in how we think about, document and frame the issues that matter to them.

First, it’s important to understand what bias is.  Basically – and I’m sure I’m butchering some definition somewhere – it’s really just substituting our expectations about a group and their shared traits, whether justified or not, for actual detailed findings about an individual in that group.  And we all do it.  It would be hard to get through life without doing it.  In many cases, it’s simply our brain’s way of coming to conclusions more quickly, although of course it can also be the result of out-and-out prejudice.

Or, as the book puts it:

That expectation is assembled from the artifacts of culture: headlines and history books, myths and statistics, encounters real and imagined, and selective interpretations of reality that confirm prior beliefs

…

Biased individuals do not see a person.  They see a person-shaped daydream

…

The individual who acts with bias engages with an expectation instead of reality

Consider the case of Ben Barres, a neurobiologist who experiences a new level of privilege after he transitions to male.  People started listening to him more carefully; he was interrupted less; his authority wasn’t questioned.  The book notes that a scientist who didn’t know he was trans was overheard saying:

Ben gave a great seminar today – but then his work is so much better than this sister’s

Funny, I guess.  But also really sad.  In other words, Ben used to be treated as a member of the class of women until he transitioned; then he was seen to be in the class of men.  His work hadn’t changed; it’s just that people placed him in a different class, which they had different expectations of.

(Side note:  I’ve been asked whether I’ve experienced any changes in how I’m treated since my transition.  The short answer is no; but then again, there’s a pandemic, and the number of new people I interact with is pretty limited.  That could all change when – or if – we ever emerge into some semblance of The Time Before The Pandemic.)

In any case, those different expectations have significant consequences.

Same-sex couples are more likely to be turned down for a home loan than a heterosexual couple; prospective graduate students whose name sounds like they are of minority races are less likely to hear back from faculty members than if their name is more Anglo sounding.  A white applicant with a criminal record is more likely to get a callback than a Black applicant without one.

Those are all egregious examples of how discrimination hurts people, but there are smaller and more constant effects as well, as the simulation in the book documents.

Yet there are ways, big and small, to hijack the brain’s normal attempt to categorize people, events and facts.  For example, in one experiment in France, they placed posters of photos of people of Arab descent with their name, age and a trait: “optimistic” or “stingy,” or something else.

People exposed to the posters were more likely to sit closer to an Arab person in a waiting room or willing to help an Arab woman who spilled her belongings. The theory here is that exposing people to the idea that Arabs were individuals and not a monolithic group helped them treat the Arabs they met on a case-by-case basis, and not as a single group.  Just as interesting:

Rather than trying to foster positive attitudes by, say, introducing French people to Arab traditions or highlighting positive role models, the poster instead emphasized how varied people are within the group identified as Arab.  Stereotypes rest on the idea that all members of a group have traits in common, and studies show that people do indeed show more bias towards individuals in a group if they seem them as homogenous. By contrast, the more we perceive that a group is composed of people who are wildly different from one another, the less we tend to stereotype.

Another experiment, in a summer school, had teachers deliberately downplay the categories of boys and girls.  So no “hello, boys and girls” or “boys sit here and girls sit there.”  Instead, they handed out yellow and blue T-shirts at random to the students, and emphasized those instead (“hello, yellows and blues” and so on.)

If teachers emphasized differences between the yellow and blue teams, biases emerged among them, with blues thinking blues were superior, and vice versa.  When they didn’t, no one cared much.  In other words, we learn what categories are important when society indicates they matter. More importantly, gender biases weren’t anywhere as prevalent as when teachers emphasized the categories of boys and girls.

In another school setting, in Sweden, teachers realized that they were the ones reinforcing gender stereotypes, for example by tolerating more noise from boys or comforting girls more. So they took pains to shift their behavior, in effect encouraging non-stereotyped gender activities.  That didn’t lead to the kids not recognizing gender, but simply having fewer preconceptions about what girls and boys are supposed to do. 

They saw boys and girls, but they stereotyped them less

But these are long-term solutions.  What can we do right now?

Checklists help.  A lot.  They slow the mind down from making quick, snap decisions, and that allows us to evaluate situations and people more carefully.

Choice architecture – a fancy way of saying how we present information and default decisions – also plays a big role.  We can engineer our way around some of our immediate biases.

Blind tests are a great way of shielding us from irrelevant information that colors our views.  I’ve written about this before, but when American orchestras moved to having “blind auditions,” the number of women hired increased dramatically.  Remember, the people judging the auditions weren’t able to actively choose women; but when they couldn’t see the candidates, orchestras somehow selected more women than when they could both see and hear who was playing.

And of course, even if we do manage to get past all our biases and hire more diverse staff, we have to make sure they get the help the need to survive, never mind excel.

To create less biased environments, it’s not enough to simply increase the diversity of the group – to add women or any underrepresented group and hope for the best. If the people who increase a group’s diversity feel devalued and unwelcome, diversity is a battle half-won When organizations fail people from marginalized groups by showing them in ways subtle and overt that they are not valued, they recruit talent only to hemorrhage it

There’s tons more in the book. You should read it. My take-aways – or at least some of them – are that we have to fix systems more than people.