Imagine a legacy passed through the ages—not wealth, but unseen burdens. This is a generational trauma, the way our unresolved anxieties shape those who follow. We see how a parent’s fears become a child’s reality. Now, imagine this dynamic reflected in artificial intelligence (AI). What if AI subtly inherits the anxieties and prejudices woven into the data used to train it?
THE GHOST IN THE MACHINE
Amazon’s experimental hiring tool, which was found to systemically discriminate against women, is a sobering reminder that our past and present biases become embedded in the technologies we build, and AI is no exception. The gender data gap means technology is often designed with male users in mind. These life-altering inequalities are hardwired into our tech.
This leads to a disturbing possibility: AI trained on data containing our anxieties and prejudices may inadvertently perpetuate—and even amplify—cycles of bias.
BIAS IN: BIAS OUT
How can we expect AI to be any different when, for example, the current gender data gap is just starting to be recognized?
Products and services are often designed with default male users in mind. According to Caroline Criado Perez, author of Invisible Women, “From cars that are 71% less safe for women than men (because they’ve been designed using a 50th-percentile male dummy), to voice-recognition technology that is 70% less likely to accurately understand women than men (because many algorithms are trained on 70% male data sets), to medication that doesn’t work when a woman is on her period (because women weren’t included in the clinical trials), we are living in a world that has been designed for men because for the most part, we haven’t been collecting data on women.”
Studies also reveal that medical algorithms designed to assess pain or determine health care eligibility can exhibit bias against Black patients.
These examples provide a stark reminder that the data we feed AI might carry inherited inequalities with enormous consequences. It’s time to acknowledge the disquieting possibility that AI, haunted by its creators’ anxieties and inherited prejudices, may inadvertently continue the cycles of generational bias and could even amplify them.
EACH GENERATION LEAVES ITS MARK
Each generation, shaped by distinct anxieties and ideals, leaves an unseen imprint on AI:
- Boomers: As children of war, their scarcity mindset might subtly shape AI decision-making.
- Hippies: Their utopian ideals might clash with the realities of social structures when encoded into AI.
- Gen X: Witnessing their parents’ struggles, they may encode a pragmatic individualism into AI, leading it to prioritize efficiency over the greater good.
- Millennials and Gen Z: Concerned with inequality and climate change, they may shape AI toward perceived fairness, possibly at the expense of individual needs.
THE TANGLED WEB OF INTERSUBJECTIVITY
This interplay of generational biases creates a complex web of intersubjectivity—the shared understanding between humans and AI. AI trained on such data may perpetuate these biases, leading to discriminatory practices in areas like loan approvals, facial recognition software, hiring algorithms, and even health care.
For example, when tech entrepreneur David Heinemeier Hansson complained to Apple that his wife was given a limit for her Apple credit card 20 times lower than his despite having a higher credit score, workers at the company informed him that it was not discrimination, it was “just the algorithm.”
Generational trauma is just one piece of the complex puzzle of AI bias. The lack of diversity in the tech sector and the relentless pursuit of profit profoundly shape how algorithms are built and the data they learn from. Understanding these interconnected factors is crucial in striving toward a more equitable and unbiased future for AI.
THE PATH TO A BETTER FUTURE
Awareness is the first step. Here’s how business leaders can break the cycle:
- Diversify the workforce: Teams with diverse ages, backgrounds, and experiences broaden the perspectives shaping AI.
- Demand transparency: Expose hidden biases by scrutinizing the origin and composition of training data.
- Explainable AI: Unravel AI’s decision-making processes to identify and address biases.
UNCOVERING OUR BLIND SPOTS
Before tackling AI bias, leaders should look inward. What values and experiences color your view of the world? How might those shape your definition of “fairness” and “ideal outcomes”?
Understanding the defining events, social movements, and technological shifts that molded each generation immerses you in a historical context. This provides a crucial baseline for identifying biases that might linger unseen within the datasets you encounter.
Always ask: “Whose voice is missing?” Remember Apple’s initial omission of period tracking in their health app—a glaring example of whose perspective was absent.
HOLD THE FUTURE TO A HIGHER STANDARD
Build a team as diverse as the world your AI seeks to navigate. Assemble developers, data scientists, and ethicists of various ages, races, genders, and socioeconomic backgrounds. If you can’t hire them, include their input.
Expand your datasets to encompass the full spectrum of lived experiences. Test sets designed to uncover hidden biases act as preventative maintenance, halting the deployment of harmful algorithms.
Integrate a “human-in-the-loop” approach for high-stakes decisions. Let AI offer recommendations, but always have a human, trained explicitly in bias awareness, make the final determination.
Finally, prioritize transparency and explainability. Document where your data comes from and strive to create AI models whose inner workings can be understood and interrogated. Foster a system of checks and balances, holding the algorithms you produce to a higher standard.
THERE’S NO FINISH LINE
Bias is an inescapable byproduct of human perception—a flaw woven into the AI fabric we create. The goal is not the impossible pursuit of a perfectly unbiased system but a relentless commitment to continuous improvement. It’s a commitment to understanding the intersubjectivity between humans and AI, knowing that the biases we grapple with today may give way to subtler, unforeseen ones tomorrow.
And this journey is never complete. Stay vigilant, embrace a critical mindset, and constantly interrogate your datasets and teams. In doing so, we move closer to a future where AI isn’t an echo of past injuries and injustices but a potentially powerful tool for building a more equitable world for all.
Jason Alan Snyder is the Global Chief Technology Officer at Momentum Worldwide.