I. “…on the basis of sex…”

“Well, I guess I’m not going to become a lawyer then,” my daughter declared after observing a mock trial I participated in. “Why would you say that?” I asked. “Because I’m not white and I’m a girl.”

Not exactly what I was going for when I invited her and my boys, in bring-your-kid-to-work day spirit.

What should trouble us all is the fact that my daughter is quite possibly the last high school student in the world who should think she can’t compete with the boys in any arena. Her academic performance borders on the absurd. And she started for her boys varsity high school basketball team and soccer teams that year,¹ serving as team cocaptain for the latter this year.

The sad reality though is she quit basketball. Hung ‘em up, giving up her chance to play for her school during her now senior year. Had nothing to do with any physical disadvantage playing with the boys. And it sure as hell didn’t have anything to do with any concerns about her “physical safety.”²

It had everything to do with the similar sort of gender discrimination that she observed during my mock trial, where a very “old boy” directed an overtly sexist critique at my co-counsel, an experienced and excellent practicing attorney. That act was rooted in a history of a sort of institutional discrimination that the legal world has successfully mitigated-in-part over the past half-century-plus but has not entirely left behind.

In comparison, algorithmic discrimination from the implementation of AI is a next-level threat. It is an institutional X-ism generating agent, if—and from all indications, when—it is or will be left to its own devices.³

II. “R-E-S-P-E-C-T. Find out what it means to me….”

“Basketball respect” from referees is defined as the benefit of the doubt for foul calls.

My daughter got no respect from the refs. I’m not suggesting this was a conspiracy. I doubt there was any malice or even consciousness on the part of any individual referee.

Basketball is one of the hardest sports to ref. It is a fast-paced, high-contact sport. You could—but of course can’t—call a foul on every single play. A lot of calls aren’t made based on what actually happens, but rather often on what the ref anticipates will happen based on whether a particular play tracks certain patterns associated with a foul call. If you look the part and put yourself in a situation where fouls are often called, you will often get the call before it happens. And if you just don’t look the part, you often won’t get the call even if it happens. A great example of this is the much-maligned bigger, slightly awkward kid who gets fouls called on them all the time just because they are standing in the vicinity of the ball, independent of making any actual contact with the ballhandler.

Hence the problem. Maddie does not and will never look the part of a boys varsity basketball player. In a sport where ref’ing is exceedingly subjective, she simply was never going to get a fair shake. Don’t believe me?

[cont’d ↗]

III. The Old-Boys’ Network 2.0: Discrimination in the Age of AI

“I just want to say that I’m not racist. I don’t even see race. Not even my own. People tell me I’m white and I believe them…”  Stephen Colbert

A. The glass ceiling endures: Progress without parity

First, if you don’t believe the “old-boys’ network” still has a presence in the practice of law or in business in general, you’re simply kidding yourself.

I wholly acknowledge and applaud the fact that there are more opportunities for women and minorities today than there were in decades past.

But based on the quality of the current public discourse on the subject matter alone, it’s clearly not the time to unfurl the “Mission Accomplished” banner just yet. The disproportionately few C-Suite members that are women and/or minorities to this day firmly establishes otherwise.⁴

B. AI is a poison for sure…

Let’s use my daughter’s basketball video to provide an admittedly simplistic illustration of the perils of relying on AI to referee a basketball game.    

If, e.g., the AI trained solely on this video and the AI were to develop its referee model focused solely on physical contact, the resulting AI ref would, amongst other things:

1. Never call a foul when a defender runs/jumps out under the shooter, not allowing the shooter to land.
2. Never call a foul when a teammate blatantly holds a defender to set a screen for a shooter.
3. Always call a foul when a defender is in the vicinity and the ballhandler falls down.
4. Never call a foul when a defender does a chasedown on a breakaway and absolutely mauls the person driving to the basket for a layup.

If, however, the AI is allowed to develop its referee model using “alternative data”—i.e., data that does not on its face have anything to do with a basketball foul but may somehow be correlative with if/when basketball fouls are called—then, well, the sky’s the limit. Taking, e.g., the last example:

4.  Never call a foul when a [male] defender does a chasedown on a breakaway and absolutely mauls the [female/long-haired/Asian/undersized/stoic, etc.] player driving to the basket for a layup.

Now, obviously this is a bit of an artificial construct, as I created the video to compile certain perceived injustices.

But while increasing the sample size of the training data used, using full game video, etc. would unquestionably regress toward the mean, what if the simple reality is that discrimination against, e.g., female, Asian basketball players by basketball referees just happens? Well, AI left to its own devices is designed to replicate the decision-making made by whatever sample of human referees it is trained on in real life. This will include by default whatever prejudices they may collectively exhibit, whether consciously or unconsciously.