At a consortium meeting this week I asked whether we have any data on whether it’s working.

Not the trainings themselves — the goal underneath them. We’ve funded anti-bias curriculum. Implicit-bias training. Black Joy programming. Years of it, across a lot of programs, with real money and real facilitator hours behind it.

My question was simple. Have we made a dent? Is there anything anyone is tracking?

The answer was no. Nothing.

I want to be careful here, because “no data” is easy to weaponize and that isn’t what I’m doing. The absence of measurement is not evidence that the work failed. It’s just an absence. But it means that everything any of us believes about whether this work is landing — including what I’m about to write — is a hunch wearing the clothes of a conclusion.

So let me at least be honest about which hunch I’m carrying.

What we count instead

We do measure things. We measure attendance. We measure completion. We count training hours and we log them in a system and we can tell you, to the person, who sat through what.

I’ve written before about why counting hours doesn’t build competency, and about the camera-off problem — the black rectangle in a required webinar that satisfies every requirement the system knows how to check. This is the same flaw pointed at a different subject. When the only thing a system can see is whether you showed up, showing up becomes the whole product.

Which leaves a question nobody is collecting evidence on: are people not getting it, or are they completing it for compliance?

I asked that out loud too. It’s not a comfortable question and I don’t think it’s a fair one to aim at providers, who are being asked to absorb this on top of licensing, enrollment, subsidy paperwork, and a business that in a lot of cases is barely holding on. If you’re in survival mode, everything arrives as one more thing to complete. That’s not a character flaw. That’s arithmetic.

But sitting with it since, I’ve started to think both of my options were wrong. Maybe the variable isn’t the curriculum, and it isn’t the provider’s sincerity either.

Maybe it’s who’s in the room.

Two rooms, same subject

I have exactly two data points, both my own, which is to say I have an anecdote and not a finding. Here they are anyway.

Years ago, at a previous workplace, race came up in a professional setting. A colleague I’d never really worked with raised it, seriously and I’m sure sincerely. It landed nowhere in me. I heard the words. Nothing moved. I remember being faintly puzzled by my own flatness — I cared about the subject, I’d have said all the right things if asked, and still the thing passed through me like weather.

Then, at the network, our former executive director Linda pulled about thirty of us onto a video call after a mass shooting. Not a training. A convening. And a Black colleague spoke — not from a framework, from her own life.

That one changed how I listened. Not for the hour. Since.

Same subject. Opposite outcome. The curriculum wasn’t the difference; there wasn’t any curriculum in the second one. The difference was that I knew her, and I didn’t know him.

And then she said the thing I’ve been turning over ever since. When something happens — when a Black person is killed — you always come to me. Why?

I recognized it immediately, from the other side of the same door. When something is AAPI, people come to me.

The design problem underneath

We run this work in affinity groups. Chinese-speaking providers in one session. Black providers in another. Spanish-speaking providers in a third. Each group gets its own space, its own facilitator, often its own language.

Everyone completes it. And nobody is ever in the room long enough to hear a story that isn’t already their own.

If relationship is the thing that makes any of this land — and my two rooms say it might be — then we have built a delivery model that systematically prevents the one ingredient that matters. That’s not a failure of any facilitator or any curriculum. That’s the design working exactly as designed, producing exactly what it was built to produce: completions.

A person who has never been in a room with you cannot be moved by a slide about you.

So the obvious next move is to put people in the same space across affinity lines, and I said as much in the meeting — that if it became a project, I’d want in.

Then I went home and ran into the problem with my own idea.

The reason it’s built this way

Affinity spaces exist for a reason, and the reason is sitting right inside my own story.

My colleague’s question — why do you always come to me? — is a description of exactly what happens when you put people in a mixed room and expect learning to occur. Someone becomes the exhibit. The burden of educating the room lands on the person with the least margin to carry it, and it lands there every single time, at no cost to anyone else in the room. Affinity grouping is, among other things, a protection against that. It’s not an oversight. It’s a considered answer to a real harm.

So the honest version of my proposal isn’t “stop doing affinity work.” It’s a narrower and much harder question: is there a design where people cross lines without anyone being conscripted into teaching?

I notice that the convening that changed me had a specific shape. Nobody was assigned to educate. My colleague chose to speak, on her own terms, in a group where she already had standing and existing relationships. Linda didn’t build a curriculum; she built an occasion, and then got out of the way. The vulnerability was offered, not extracted.

That’s a fragile thing to try to reproduce on purpose. It’s easy to do badly. Done badly, it’s worse than the silos — it’s the silos plus a performance requirement.

I don’t know how to systematize it. I’m not sure it can be systematized, which may be the actual finding here and not a failure of my imagination.

What I’d want measured

Here’s where I land, provisionally, and I hold it loosely.

The next question shouldn’t be “should the training be mixed or separate.” It should be: what are we even trying to observe? Right now the answer is attendance, because attendance is what the system can see. If we care about whether anything moved, we have to be willing to name what “moved” would look like — in enrollment patterns, in who providers say yes to, in what gets said in a room when the facilitator leaves.

That’s harder to count than hours. It’s also the only version worth counting.

And I’d want to ask the people this is aimed at. The clearest thing anyone said in that whole meeting, on a different topic, was that the voice missing from the conversation belonged to the people who’d be on the receiving end of it. That was true of the monitoring discussion. It’s true of this one too. Nobody in the room where I asked my question was going to be the one sitting in the training.

Thinking out loud

I’m publishing this unfinished on purpose.

I don’t have a conclusion, I have a tension: the story that makes me want cross-affinity work is the same story that explains why we don’t do it. Both readings are supported by the same forty seconds of my colleague talking. That’s usually a sign the question is real and not rhetorical.

What I do believe, and have written before in other contexts, is that belonging isn’t soft — it’s infrastructure. Clarity, access, predictable process, respectful tone. I’d now add proximity to someone whose experience isn’t yours, with the caveat that proximity without consent is just extraction with better branding.

If you’ve run something that crossed those lines and worked — or crossed them and went badly, which is more useful — I’d genuinely like to hear how it was built. Reach out. I’m not selling a model. I’m trying to figure out whether the one we have was ever tested.