As AI Gets Smarter, Human Irrationality Matters More
Behavioral AI
Why I’m Building Behavioral AI Lab
Hello, and welcome.
I’m Tomomi Tanaka, a researcher and practitioner working at the intersection of behavioral science, AI, and digital safety.
I began as an Assistant Professor of Economics at Arizona State University, then worked at the World Bank, Uber, Match Group, and Amazon, applying behavioral insights, data science, and AI to real-world challenges.
Today, I advise BEworks, a global behavioral science consultancy, while building Behavioral AI Lab—a new venture exploring what happens when behavioral science meets AI. Through this Journal, I'll share that journey, along with research, insights, and practical lessons from the field.
Data Can Be Right. People Still Don’t Act.
When I was analyzing education policy in Ghana, the evidence was clear: staying in school significantly increased future earnings. Strong data. Rigorous research. And yet, children kept dropping out.
The data was correct. People simply didn’t behave the way the models predicted.
That realization stayed with me throughout my career.
No matter how accurate our models become, humans remain the ultimate decision-makers.
And humans are not perfectly rational.
Technology Evolves. Human Psychology Doesn’t.
While leading safety initiatives at Uber and Match Group—the parent company of Tinder, Hinge and more than 15 dating platforms—I encountered the same challenge from a different angle.
Why do people fall for scams?
Why do users ignore warnings?
Why do people trust individuals who pose obvious risks?
These are not merely technology problems.
They are also human behavior problems.
You can build sophisticated safety systems, deploy advanced machine learning models, and design elegant user experiences. But if you ignore how people actually think, decide, and behave, truly safe systems remain out of reach.
The same principle applies to AI.
It’s Not “What Can AI Do?”—It’s “How Will Humans Use It?”
AI capabilities are advancing at an extraordinary pace.
Large language models can now reason, write, code, summarize, and generate content at a level that seemed impossible only a few years ago.
But I’d argue something different:
The smarter AI becomes, the more critical human understanding becomes.
People may overtrust AI systems and rely on them when they shouldn’t.
Organizations may struggle to adopt AI even when the technology is valuable.
Bad actors may exploit AI systems for manipulation, fraud, and harm.
Ultimately, whether AI succeeds or fails depends not only on the technology itself, but on human behavior.
That’s why I don’t think of AI as merely a technology. I think of it as a social system that interacts with human decision-making.
A Behavioral AI Perspective
Integrating behavioral science into AI design can help create systems that are safer, more trustworthy, and more aligned with human needs.
That’s the core hypothesis behind Behavioral AI Lab—and the thread running through everything I do.
In this Journal, I’ll explore:
Behavioral Science and AI Adoption — why people embrace or resist AI systems
AI Safety and Trust — how human behavior shapes AI risk
Human-Centered AI Design — building systems that work with human psychology, not against it
Digital Trust and Decision-Making — lessons from Trust & Safety in the real world
Research and Real-World Applications — bridging academic insights and practice
Because if we want to understand AI, we must first understand the humans who use it.
Behavioral AI Lab exists because I believe one of the most important challenges of the AI era is not simply building better models, but better understanding the people who interact with them.
If these topics resonate with you, I hope you'll continue following Behavioral AI Lab as we explore the human side of AI together.
What’s Next?
In my next post, drawing on a recent Harvard Business Review article, I’ll explore Safety by Design—why safety must be built into products from the very beginning, rather than bolted on as an afterthought.
Until next time,
Tomomi Tanaka
Founder, Behavioral AI Lab
BLOG & INSIGHTS
Exploring the Human Side
of Artificial Intelligence.
As AI Gets Smarter, Human Irrationality Matters More
Behavioral AI
Why I’m Building Behavioral AI Lab
Hello, and welcome.
I’m Tomomi Tanaka, a researcher and practitioner working at the intersection of behavioral science, AI, and digital safety.
I began as an Assistant Professor of Economics at Arizona State University, then worked at the World Bank, Uber, Match Group, and Amazon, applying behavioral insights, data science, and AI to real-world challenges.
Today, I advise BEworks, a global behavioral science consultancy, while building Behavioral AI Lab—a new venture exploring what happens when behavioral science meets AI. Through this Journal, I'll share that journey, along with research, insights, and practical lessons from the field.
Data Can Be Right. People Still Don’t Act.
When I was analyzing education policy in Ghana, the evidence was clear: staying in school significantly increased future earnings. Strong data. Rigorous research. And yet, children kept dropping out.
The data was correct. People simply didn’t behave the way the models predicted.
That realization stayed with me throughout my career.
No matter how accurate our models become, humans remain the ultimate decision-makers.
And humans are not perfectly rational.
Technology Evolves. Human Psychology Doesn’t.
While leading safety initiatives at Uber and Match Group—the parent company of Tinder, Hinge and more than 15 dating platforms—I encountered the same challenge from a different angle.
Why do people fall for scams?
Why do users ignore warnings?
Why do people trust individuals who pose obvious risks?
These are not merely technology problems.
They are also human behavior problems.
You can build sophisticated safety systems, deploy advanced machine learning models, and design elegant user experiences. But if you ignore how people actually think, decide, and behave, truly safe systems remain out of reach.
The same principle applies to AI.
It’s Not “What Can AI Do?”—It’s “How Will Humans Use It?”
AI capabilities are advancing at an extraordinary pace.
Large language models can now reason, write, code, summarize, and generate content at a level that seemed impossible only a few years ago.
But I’d argue something different:
The smarter AI becomes, the more critical human understanding becomes.
People may overtrust AI systems and rely on them when they shouldn’t.
Organizations may struggle to adopt AI even when the technology is valuable.
Bad actors may exploit AI systems for manipulation, fraud, and harm.
Ultimately, whether AI succeeds or fails depends not only on the technology itself, but on human behavior.
That’s why I don’t think of AI as merely a technology. I think of it as a social system that interacts with human decision-making.
A Behavioral AI Perspective
Integrating behavioral science into AI design can help create systems that are safer, more trustworthy, and more aligned with human needs.
That’s the core hypothesis behind Behavioral AI Lab—and the thread running through everything I do.
In this Journal, I’ll explore:
Behavioral Science and AI Adoption — why people embrace or resist AI systems
AI Safety and Trust — how human behavior shapes AI risk
Human-Centered AI Design — building systems that work with human psychology, not against it
Digital Trust and Decision-Making — lessons from Trust & Safety in the real world
Research and Real-World Applications — bridging academic insights and practice
Because if we want to understand AI, we must first understand the humans who use it.
Behavioral AI Lab exists because I believe one of the most important challenges of the AI era is not simply building better models, but better understanding the people who interact with them.
If these topics resonate with you, I hope you'll continue following Behavioral AI Lab as we explore the human side of AI together.
What’s Next?
In my next post, drawing on a recent Harvard Business Review article, I’ll explore Safety by Design—why safety must be built into products from the very beginning, rather than bolted on as an afterthought.
Until next time,
Tomomi Tanaka
Founder, Behavioral AI Lab
BLOG & INSIGHTS