When I first started working with large language models, I was mesmerized. Here were machines that could write poetry, translate languages, even offer advice on moral dilemmas. But the more I interacted with them, the more something started to gnaw at me—not because they made mistakes, but because they often made sense only within a narrow cultural frame. They could generate eloquent responses, yet lacked the context to know when silence might be more respectful than speech, or when indirectness is a form of courtesy, not evasion.
It was in those moments that I realized: technical alignment is not enough. Cultural alignment is essential.
The Missing Layer in AI Alignment
AI alignment often refers to designing systems that act in accordance with human goals and values. But “human values” are anything but monolithic. What matters in one culture might be meaningless—or even offensive—in another. For example, direct disagreement may be encouraged in some societies and avoided in others. A language model trained to “speak truth to power” may need to unlearn that instinct in a collectivist environment where face-saving and group harmony are valued above individual assertiveness.
These aren’t edge cases—they are the fabric of global human interaction.
Designing AI That Understands Us (All of Us)
This is what drew me to the intersection of AI and cultural psychology. In my recent projects, including the development of a Fatwa recommendation engine and a regionally aware legal assistant, I’ve seen firsthand how crucial cultural norms are in shaping human-AI interaction. What counts as “helpful advice” in one setting can be perceived as intrusive or disrespectful in another. And when trust is at stake—especially in legal, ethical, or emotional domains—misalignment can be more than a glitch. It can cause real harm.
Building culturally aligned AI isn’t just about training data or language variation. It’s about designing models that respect the moral, social, and psychological nuances of the people they serve.
Lessons From Psychology, Not Just Programming
The exciting part is that we don’t have to build this from scratch. Cultural psychologists have been studying how people across societies reason, make decisions, and resolve conflicts for decades. Their insights offer a roadmap for what AI systems should consider when navigating culturally sensitive situations—how different societies conceptualize fairness, authority, obligation, or autonomy.
The real challenge—and the opportunity—is integrating these insights into AI design and evaluation. Can we map cultural dimensions (like Hofstede’s or Schwartz’s) into model objectives? Can we design prompt structures that adapt to social context, not just user input? These are the kinds of questions that inspire me, and the ones I hope to explore more deeply in collaborative, interdisciplinary environments.
Toward a More Inclusive AI Future
We are at a critical point in the evolution of AI. As systems become more powerful and ubiquitous, the risks of cultural misalignment multiply. But so do the opportunities. By bringing together computer scientists, psychologists, ethicists, and community voices, we can build AI that doesn’t just mimic intelligence—it demonstrates wisdom.
I believe the future of AI lies not only in better algorithms, but in deeper empathy—and that starts by asking whose values we’re encoding, whose voices we’re prioritizing, and whose context we’re honoring.
That’s the kind of AI I want to help build.
