Awareness is not a control surface
Why organizations can express the right values for a decade and produce the same promotion numbers every year.
LaToya Bowlah’s piece in Fast Company names something real: organizations have gotten good at performing awareness of gender inequity while producing nearly identical advancement outcomes year over year. She calls the mechanism an Awareness–Accountability Gap, and she’s right about how it works. Leaders earn credit for expressing the right values. That credit lowers scrutiny. Scrutiny reduction removes pressure for action. The loop closes, and nothing changes.
I want to add a layer she gestures at but doesn’t fully build out, because it’s the layer that matters most to the technical leaders I write for.
The empathy ceiling isn’t just a cultural failure. It’s an information failure. And the reason it’s expensive is the same reason any information failure is expensive: decisions get made on inputs that don’t reflect reality.
What this looks like in an engineering organization
A senior woman engineer pushes back on a launch timeline. She names a specific reliability risk, proposes a concrete alternative, and frames it as a business decision, not a preference. The room hears it as friction. Her manager, who considers himself a genuine ally, interprets his discomfort as a signal about her style rather than as information about the decision. He lets the launch proceed. The timeline slips. The risk she named materializes. In the postmortem, the failure gets attributed to execution.
She did the job. The information didn’t travel. The organization learned nothing.
This is what Bowlah is describing when she talks about “ambiguity transfer.” In technical organizations, that cleanup is often performed by the people with the most accurate read on the system and the least formal authority to make the call. The cost isn’t just unfairness. It’s degraded signal quality at exactly the points where the signal matters most. I’ve written before about how the likability tax produces exactly this outcome: the same boundary-shaped behavior gets read as “decisive” from one leader and “obstructive” from another, and the misread gets coded as a performance observation rather than as a systems failure.
The psychology underneath it
Social psychologists call this moral licensing. The phenomenon was documented in a well-known 2001 study by Benoit Monin and Dale Miller: people who had established their credentials as non-prejudiced were subsequently more willing to make decisions that disadvantaged women and minority candidates, not less. Their prior virtue functioned as a credit balance against which bias could be quietly drawn. (A 2015 meta-analysis of 91 studies found the overall effect is small to medium in size, not large, which is worth knowing. But the mechanism itself is well-established across the broader literature.)
In organizational terms: the manager who attended the DEI training, who considers himself a girl dad, who uses the right language in calibration meetings, has implicitly offset some of his accountability. He doesn’t experience himself as the problem. He’s one of the good ones. The license is invisible to him.
This is not a character critique. It’s a systems observation. Moral licensing isn’t a flaw in bad people. It’s a predictable mechanism in all people. The response isn’t to find less biased managers. It’s to build processes that don’t rely on managers noticing their own licenses. This is the same logic I’ve applied to the unexamined leader as an operational cost: you don’t solve the problem by asking the system’s weakest signal (the leader’s self-perception) to become its strongest.
Why the standard interventions fail under pressure
Bowlah’s four interventions (track advancement velocity, stop rewarding awareness alone, make sponsorship visible, assign ownership to ambiguity) are correct. The missing piece is how to build them so they function under pressure rather than just on paper.
The pattern I’ve seen is this: organizations implement the right practices during periods of low stress and abandon them when stakes are high. The calibration rubric gets used in Q3 planning but quietly set aside when the VP needs to staff the critical project fast. The sponsorship tracking happens until the company misses a quarter. The accountability mechanism works until it’s inconvenient.
This is what I mean when I say the control surface has to be structural, not conversational. A norm is not a control surface. A stated commitment is not a control surface. A process that has to be actively circumvented, that creates friction when it’s bypassed, that leaves a visible record when it’s skipped, that’s a control surface.
Concretely: the “Decision Transcript” check I’ve described in the likability tax piece is one version of this. When feedback about a woman leader uses evaluative labels without content (”difficult,” “not collaborative,” “abrasive”), require five things before the label can influence a performance narrative: what decision was on the table, what she recommended, what risk she named, what alternative she offered, what the business chose and what happened next. If you can’t produce the transcript, you don’t have a performance observation. You have a vibe.
Applied to Bowlah’s sponsorship problem: the same logic holds. If a leader is sponsoring someone, document it. Who got the visibility? Who got the room? Who got credited when the outcome landed? Sponsorship that leaves no record isn’t sponsorship. It’s social warmth, and social warmth doesn’t transfer into promotion decisions. McKinsey’s 2024 Women in the Workplace data shows this is where the pipeline breaks down first: the “broken rung” from entry to manager, where 81 women are promoted for every 100 men, is largely a sponsorship and visibility problem dressed up as a performance problem.
The mechanism Bowlah doesn’t name
There’s one more mechanism that deserves airtime in technical organizations specifically.
Status characteristics theory, developed by Joseph Berger and colleagues and extended significantly by Stanford sociologist Cecilia Ridgeway, explains why mixed-status groups systematically disadvantage people who hold low-status characteristics, even in domains where status is irrelevant to competence. Ridgeway’s Framed Before We Know It is the most direct articulation of how this operates in organizational settings: gender acts as a background identity that biases the performance of behaviors undertaken in the name of role and expertise, without anyone involved necessarily intending it. In engineering rooms specifically, gender operates as what the theory calls a diffuse status characteristic: it shapes who gets treated as competent before any evidence has been presented. The effects are automatic and largely unconscious. They show up in who gets interrupted, whose proposals get credited, whose concerns get escalated, and, critically, whose technical objections get treated as data rather than friction.
The relevance here is that empathy doesn’t touch this mechanism. A manager can be deeply committed to equity, feel real sympathy for his colleague’s experience, and still unconsciously downweight her risk assessment in the design review. The feeling and the behavior run on different tracks.
What does touch it is expectation-setting: explicitly establishing technical credibility before group interaction, distributing air time structurally instead of leaving it to organic norms, and recording whose concerns were raised and whether they were acted on. Process interventions like these work because they don’t ask the unconscious to behave differently. They change what the situation makes easy.
The engineering leadership version
The Fast Company piece ends with the observation that until awareness is paired with measurable, tracked, consequential accountability, inequity will persist behind the language of progress.
That’s the argument in one sentence, and it’s the right one.
The engineering leadership version is this: your organization is a system, and systems produce the outcomes they’re designed to produce. Right now, most technical organizations are designed to reward the performance of equity over its execution. The design wasn’t anyone’s intention; it emerged from the incentives. But it’s legible, it’s measurable, and it can be changed.
The question is whether you’re willing to build processes that create friction when they’re bypassed, processes that inconvenience the people who currently benefit from how the system runs. That’s what accountability actually requires. Not a better stated value. A harder-to-bypass mechanism.
Empathy that costs nothing produces nothing. The control surface is where the work lives.