How Politics Is Stalling The Fight Against Biased AI

Tech companies are facing a new wave of scrutiny, not for their workplace diversity programs, but for their efforts to address bias in artificial intelligence products. After making strides to improve diversity, equity, and inclusion, the industry is now navigating a political landscape that frames these efforts as a partisan issue.

The Political Assault on Woke AI

The term “woke AI” has gained traction in Washington, shifting the focus from fixing harmful algorithmic discrimination to investigating what some lawmakers call ideological bias. In March, the House Judiciary Committee issued subpoenas to major tech players including Amazon, Google, Meta, Microsoft, and OpenAI. The investigation aims to determine if the Biden administration pressured these companies to censor speech or advance a specific agenda through their AI.

This political shift is also evident in government agencies. The U.S. Commerce Department's standard-setting branch has notably removed language about AI fairness, safety, and “responsible AI” from its research initiatives. Instead, it now calls for a focus on “reducing ideological bias” to promote “human flourishing and economic competitiveness.”

A Chilling Effect on Scientific Progress

Researchers in the field are sounding the alarm. They warn that this partisan pressure could create a chilling effect on scientific inquiry, making developers hesitant to address known technical flaws in AI for fear of political backlash. The concern is that legitimate issues of algorithmic reliability, accuracy, and safety will be sidelined, hindering innovation while leaving fundamental bias problems unsolved.

For many in the tech world, this represents a significant shift in priorities driven by Washington. Ellis Monk, a Harvard University sociologist, experienced the previous era firsthand. Google had sought his expertise to make its AI products more inclusive, particularly in computer vision, which often struggled to accurately represent people with darker skin tones.

A Tangible Example The Monk Skin Tone Scale

Monk, a scholar of colorism, developed the Monk Skin Tone Scale, a more inclusive color scale that Google adopted to improve how its AI image tools portray human diversity. This replaced an outdated standard designed primarily for white dermatology patients. The change was a success, with Monk noting that “consumers definitely had a huge positive response.”

While he believes the scale itself is secure because it's integrated into numerous products, he worries about the future. “Could future funding for those kinds of projects be lowered? Absolutely, when the political mood shifts and when there’s a lot of pressure to get to market very quickly,” Monk stated.

A History of Algorithmic Bias

Before the current political debate, a wealth of research highlighted the real-world harms of AI bias. Studies showed that self-driving car technology had difficulty detecting darker-skinned pedestrians. AI image generators asked to create a picture of a surgeon overwhelmingly produced images of white men. Facial recognition software has misidentified Asian faces and led to the wrongful arrests of Black men. Even a decade ago, Google's own photo app infamously labeled a picture of two Black people as “gorillas.”

The Gemini Incident A Catalyst for Controversy

Efforts to combat these biases led to the controversy surrounding Google's Gemini AI chatbot. To prevent the tool from perpetuating stereotypes, Google implemented technical guardrails. However, the system overcompensated, generating historically inaccurate images, such as depicting America's founding fathers as Black, Asian, and Native American men. Google quickly apologized and paused the feature, but the incident became a powerful symbol for critics of “woke AI.”

Vice President JD Vance referenced the event at an AI summit, condemning the advancement of “ahistorical social agendas through AI” and promising that a Trump administration would ensure AI systems are free from ideological bias.

Two Sides of the Same Coin

Alondra Nelson, a former Biden science adviser, sees a strange overlap in the rhetoric. She argues that labeling AI as “ideologically biased” is fundamentally an admission of the very problem of algorithmic bias that experts have been working to solve for years. However, she is not optimistic about collaboration. “Problems that have been differently named — algorithmic discrimination or algorithmic bias on the one hand, and ideological bias on the other — will be regrettably seen as two different problems,” Nelson said. Amid the current political climate, finding common ground appears unlikely.