Who Programmed the Programmer?

At the HBCU where I teach, we are explicitly told to teach APA style to our students.

Then I read APA’s apology for helping promote and perpetuate racism.

Sit with that for a moment. An organization that acknowledged harming Black communities also supplies writing standards we are expected to pass on to Black students. I teach those standards. I also want us to examine the authority behind them.

As a media professor, creator, and app developer, I keep returning to a question that connects my classroom to the screen in your hand: Who gets to decide what “correct” looks like?

We call television content programming. We call the instructions behind software programming. We also use the word to describe learned habits and assumptions. Those meanings differ, but each invites us to examine who makes the choices and whose perspective becomes the default.

I have put that question to an AI assistant, too. Before I accept its answers, I want to understand the human judgments behind them. Who programmed the programmer?

As a Black man, I want people who understand experiences like mine involved in those decisions. I also know that sharing my skin color does not guarantee sharing my perspective. A Black developer can inherit and reproduce biased ideas about intelligence, professionalism, beauty, or credibility.

Representation matters. So does the thinking we bring into the room. And so does having enough authority to change what happens there.

The American Psychological Association is a scientific and professional organization representing psychology in the United States, with influence over research, education, and practice. Founded in July 1892, it issued its formal apology in October 2021, 129 years later. It acknowledged harms involving research, testing, and the treatment of people of color. Generations of professional authority preceded that acknowledgment.

APA itself says its writing style is used by millions of people worldwide. It reaches student papers, dissertations, and published research. Its guidance covers how writers organize information, cite evidence, and describe people. When a school or publication requires those conventions, an association’s guidance becomes a requirement in someone else’s classroom or career.

Using APA Style does not endorse APA’s history. It does make the organization’s authority part of everyday academic life. We should be able to teach a useful standard and question the institution behind it.

The contradiction carries weight because APA’s stated mission is to benefit society and improve lives. Its current Ethics Code includes doing good and avoiding harm among its aspirational principles. Who gets to define the good when the people being helped say they are being harmed? A promise to help cannot substitute for their ability to challenge the institution’s judgment.

The Association of Black Psychologists rejected the apology as insufficient. Its critique challenged the historical account and called for independent authority for ethnic psychological associations over areas including ethics, education, and training.

Fairness also requires acknowledging the work APA reports doing since the apology. Its 2024 progress report describes recurring meetings with leaders of ethnic psychological associations, financial support for their conferences, and changes intended to respect their independence. It also describes a racial and ethnic diversity task force advising APA’s leadership and new reporting standards for discussing race, ethnicity, and culture in research. Those are concrete efforts, and they deserve to be recognized.

I can recognize that work and still ask how far it reaches. Consultation creates an opportunity for influence; shared authority determines who can change the decision. Participation does not automatically mean endorsement. Reform has to reach inherited practices wherever they persist, and the people affected need a meaningful role in judging the results.

And here is the irony: APA’s efforts to repair exclusion have collided with a federal push against diversity initiatives. President Trump’s January 2025 executive order framed that push as protecting civil rights and restoring merit. By March, the conflict had reached psychology’s training infrastructure. APA’s operationally independent Commission on Accreditation temporarily suspended evaluation of certain diversity standards concerning recruitment, admissions or selection, and retention. APA attributed the action to the administration’s legal interpretation and developments surrounding the order.

That puts another question on the table: Who gets to define merit, and how do we know the definition has accounted for the exclusion APA itself acknowledged? I can demand more from APA’s reforms while opposing pressure that weakens efforts to address those harms. An institution’s overdue reckoning should be a beginning, not the moment we decide the work has gone too far.

An apology does not rewrite an algorithm.

It does not automatically revise a newsroom’s habits, a university’s requirements, or the data used to train a model. That is where this story moves from an institution’s past into our daily lives.

Imagine two stories about people doing the same thing. One headline calls a person “assertive.” Another calls a person “aggressive.” The facts may be similar, but the labels invite different judgments.

Now imagine those choices repeated across years of stories, scripts, images, and commentary. I want my students to notice who gets to be complicated, who gets to be an expert, and who keeps appearing as a problem somebody else must solve.

Audiences can question and resist those messages. On social media, though, we exercise our choices inside systems that also make choices about us. We choose whom to follow and what to post. The platform decides how content is ranked, recommended, and moderated. Freedom to speak and control over distribution are different forms of power.

APA’s own 2023 advisory on adolescent social media use explicitly acknowledges that social-media algorithms can encode historical racism and discrimination, sometimes reflecting the perspectives of their builders. It also recognizes that users’ experiences are shaped by both their choices and platform features they may not see. That acknowledgment brings us back to the programmer and the world that shaped the programmer.

A September 2023 article in APA’s Monitor on Psychology reported criticism of TikTok’s appearance-based account recommendations for their segregating effect. It also described beauty filters on Instagram and Snapchat that could lighten skin or apply more typically White facial features. These are reported examples from 2023, rather than an audit of today’s platforms.

Consider the message in a filter that presents lighter skin as an improvement. A beauty judgment has become a product feature. The user can choose to apply it, but somebody else has already defined what the transformation should look like. That is where my question about programming becomes visible on the screen.

Digital redlining takes the concern beyond representation. The term describes ways digital systems reproduce discriminatory barriers to information, services, or opportunity. A personalized feed is not automatically redlining. The question is whether the system systematically excludes people along racial or other protected lines. Being free to open an app means little if opportunities are quietly withheld from what you get to see.

Housing advertising provides a concrete example. In a 2022 case against Meta, the Justice Department alleged that Facebook’s housing-ad system discriminated on grounds including race, and that its delivery algorithms relied partly on protected characteristics. Under the court-approved settlement, Meta discontinued its Special Ad Audience tool for housing ads and developed a system to reduce disparities in ad delivery, with independent compliance review.

The settlement shows what repair can require: changes to the system, measurable targets, and outside scrutiny. Civil-rights enforcement reached the machinery deciding who saw a housing opportunity.

Think of the square peg and the round hole. Neither shape is inherently defective. But make the circle the universal standard, and the square becomes a problem to correct. A platform that treats one community’s expression as the default deserves scrutiny over how it judges everyone else, especially when those communities had little say in the standard.

APA is one influential institution in a larger system of companies, governments, professions, and schools. Its apology gives us a reason to examine inherited assumptions across that system. AI adds another question: What happens when a machine learns those assumptions?

A 2024 study in Nature found that tested language models exhibited prejudice toward speakers of African American English, even when race was not explicitly identified. In experimental tasks, the models associated those speakers with less prestigious jobs and made harsher hypothetical criminal-justice judgments.

These were controlled tests of particular models, not a report of actual hiring or sentencing decisions. But the finding exposes a serious problem: a system can respond politely to a direct question about race while treating language associated with Black people differently.

Imagine being told a tool is fair because it knows the right thing to say about equality. I would also want to know what it does when it encounters the way my community speaks.

Psychology’s relationship with technology is concrete, too. APA Labs, part of APA Services, works with ORCHA to evaluate digital mental and behavioral health products against criteria involving evidence, safety, privacy, security, and accessibility, including AI considerations. Its badge signifies alignment with program criteria; it is not a guarantee of effectiveness or legal compliance.

Professional judgment is already part of technology evaluation. The people setting those standards, and the communities able to challenge them, matter.

These examples do not establish that APA caused a particular model’s bias. Together, they make a case for examining both the technology and the institutions we trust to judge it.

For HBCUs, this should be a call to build and to set standards.

Computer science students can work with communications students, psychologists, historians, and community members to examine the same tool. Each group may notice something the others miss. Together, they can test what a system assumes, whose language it handles poorly, and how a mistake affects an actual person.

That work deserves funding, ownership, and decision-making authority. A community should have a role in deciding what gets built, how it is evaluated, and whether it serves them.

For educators, that means teaching students how to use a standard and how to examine it. For developers and platforms, it means testing outcomes across communities, publishing meaningful findings, and providing ways to challenge harmful decisions. For anyone purchasing these tools, it means asking whose experience informed the design and what evidence supports the promise of fairness.

I use AI. I build with it. I teach students who will help shape what comes next. That puts responsibility on me, too.

We should ask who programmed the programmer. Then we should insist on a meaningful role in deciding what gets programmed next.

The square peg should have a say in the shape of the opening.