In an era defined by the relentless expansion of artificial intelligence and the growing influence of data-driven decision-making, the ability to measure society accurately has never been more critical—or more complicated. While massive datasets and machine learning models promise unprecedented insights into human behavior, they also introduce new layers of statistical complexity and potential bias. At the intersection of this methodological frontier stands Naoki Egami, an associate professor in MIT’s Department of Political Science, whose research is reshaping how social scientists approach the fundamental challenge of quantification. His work represents a synthesis of two often-disparate worlds: rigorous statistical theory and the messy, empirical realities of political and social life. By insisting that researchers must be fluent in both the technical mechanics of measurement and the substantive nuances of the problems they study, Egami is building a framework that promises not just better data, but a more accurate understanding of the world itself.
What Does It Mean to Measure Society Accurately?
Measuring society accurately is the process of using statistical models and empirical methods to draw valid, reliable conclusions about human behavior, political dynamics, and social structures. In the context of Naoki Egami’s work, this means developing methodological tools that account for the complexities introduced by artificial intelligence, sampling biases, and the inherent limitations of observational data. It is NOT simply about collecting more data; it is about ensuring that the inferences drawn from that data genuinely reflect the underlying social reality, rather than the artifacts of how the data was generated or processed.
For researchers, policymakers, and industry analysts alike, the stakes could not be higher. If the statistical foundations are flawed, every subsequent decision built upon that data—from public policy to corporate strategy—is compromised. Egami’s contribution is to provide the technical scaffolding that ensures these foundations hold, even when the analytical tools themselves are as unpredictable as human beings.
The Man Who Bridged Two Worlds
Egami’s journey to the forefront of political methodology is a story of serendipity, intellectual curiosity, and a stubborn refusal to accept disciplinary boundaries. His path did not begin with a clear vision of academia, but rather with a fortuitous sequence of events that illustrated the importance of being open to unexpected opportunities—a theme that echoes throughout his research on external validity and causal inference.
Growing up in Tokyo, Egami excelled in mathematics and physics, but found himself equally drawn to political philosophy. For a young student, this seemed like an irreconcilable conflict: the cold, deterministic logic of the hard sciences versus the fluid, contested ideas of political thought. The University of Tokyo, where he eventually enrolled, offered little immediate clarity. It was only when he wandered into a workshop about U.S. graduate programs—mistakenly believing it was about MBA degrees—that the pieces began to fall into place.
That workshop changed the trajectory of his career. A panelist, a political science PhD student, presented the idea of using mathematical tools to answer political questions. Intrigued, Egami asked a question. The miracle, as he describes it, happened after the event: the panelist descended from the stage, found him in the crowd of 200, and invited him to what he assumed was another career seminar. It turned out to be an academic talk attended by nineteen professors and one first-year undergraduate—Egami himself.
The speaker that day was Kosuke Imai, a name now synonymous with modern political methodology. Imai’s presentation on using statistics in social science was a revelation. “I was super-excited and felt if I could do even 20 percent of that, it would be a dream,” Egami recalls. He approached Imai, declared his ambition, and set about turning that ambition into reality. After a year as an exchange student at the University of Michigan, he applied to graduate schools and landed at Princeton University, where Imai eventually became one of his advisors. Working alongside luminaries like Rafaela Dancygier and Brandon Stewart, Egami began producing research on topics like external validity—a measure of whether findings from one context can be generalized to another—and found a receptive audience almost immediately.
From Princeton to Columbia, Then MIT
After completing his PhD at Princeton in 2020, Egami joined the faculty at Columbia University, where he spent five years building his portfolio. In 2025, he made the move to MIT, joining the Department of Political Science as an associate professor with tenure. He also became a faculty affiliate of the Statistics and Data Science Center at the Institute for Data, Systems, and Society (IDSS). This dual affiliation is deliberate, reflecting his conviction that political science and data science must be pursued in concert, not in isolation.
At MIT, Egami found an environment that amplifies his approach. “You need both perspectives,” he says. “If you only think about technical statistical theories, you might not work on interesting empirical problems sometimes. But if you only think about problems, you won’t really solve them optimally; you’ll solve them in an ad-hoc way. So, you really want to have both lenses.”
How Researchers Can Account for AI Bias in Social Science
One of the most pressing questions facing modern social scientists is how to integrate artificial intelligence tools into their research without compromising accuracy. As AI systems become more sophisticated, they are increasingly used to classify text, identify patterns in large datasets, and even simulate human responses. But these tools are not neutral instruments; they carry their own biases, limitations, and tendencies that can distort research outcomes if not carefully managed.
Egami’s research addresses this challenge head-on by developing frameworks to detect, measure, and correct for AI-specific errors. The core issue is that large language models and other machine-learning systems operate on probabilistic principles, producing outputs that are often indistinguishable from—but not equivalent to—human reasoning. When researchers use these tools to, say, code political speeches or analyze social media sentiment, they inherit the AI’s analytical “tendencies,” which may align with ground truth only imperfectly.
The solution, Egami argues, is not to abandon AI but to treat it as an object of study in its own right. Researchers must develop a statistical theory of AI behavior, understanding not just what the model outputs, but why it outputs that way. This involves designing validation protocols, sensitivity analyses, and calibration techniques that allow researchers to separate signal from noise. It requires a shift in mindset: instead of asking “What does this AI say?” researchers must ask “How reliable is this AI’s output for this specific task, in this specific context?”
The Pre-AI Roots of a Modern Problem
Egami’s interest in AI accountability predates the current boom. Years before generative models like ChatGPT entered the public consciousness, he was already asking uncomfortable questions about what happens when these tools are introduced into research settings. His early work laid the groundwork for a growing subfield that now commands intense attention from funding agencies, academic departments, and private industry alike.
This foresight is characteristic of Egami’s broader research philosophy. He identifies problems that are not yet on the radar of the mainstream field but are guaranteed to become significant, and he equips himself with the theoretical tools to address them. His research on external validity—the question of whether an experimental finding in one context holds in another—similarly anticipated a growing crisis in social science, where replication failures have called into question the reliability of published research.
Why Both Statistical and Substantive Knowledge Are Essential
In a landscape teeming with data, it can be tempting to treat analysis as a purely technical exercise. But Egami cautions against this view. The most sophisticated statistical model in the world will produce misleading results if applied to poorly understood empirical problems. Conversely, deep substantive knowledge without methodological rigor leads to ad-hoc solutions that cannot withstand scrutiny.
This dual-lens approach is evident in every facet of Egami’s work. He does not simply ask “What method should I use?” but rather “What is the underlying social mechanism generating this data, and how can I design a method that respects that mechanism?” By maintaining this dual perspective, he ensures that his methodological innovations are always grounded in the messiness of real-world political and social processes.
The practical implications are substantial. Policy decisions grounded in flawed statistical inference can waste billions of dollars and deepen inequality. Marketing strategies built on biased data misallocate resources. Public health interventions designed using non-representative samples may fail precisely among the populations they were intended to help. By insisting on a marriage of technical sophistication and substantive insight, Egami’s work provides a roadmap for researchers who want their findings to actually matter.
The Role of External Validity in Accurate Measurement
One of Egami’s most significant contributions has been in the domain of external validity—a concept that addresses a fundamental question in social science: are the cause-and-effect relationships identified in a study applicable to other populations, settings, and time periods? Traditional research designs, particularly randomized controlled trials, excel at establishing internal validity (whether a treatment worked in the study sample) but often struggle with external validity (whether it will work broadly).
Egami’s methodological contributions provide systematic frameworks for assessing generalizability, allowing researchers to move beyond anecdotal assumptions about whether findings apply elsewhere. This is particularly important in an era of globalized policy making, where interventions designed in one cultural or economic context are routinely transplanted into others. Without rigorous external validity checks, such transplants are effectively shots in the dark.
Different Perspectives on the Same Problem
Egami’s research interests extend beyond the narrow confines of any single methodology. He is deeply engaged with fundamental questions about politics—issues of power, representation, identity, and institutional design. These interests keep him grounded, ensuring that the technical machinery he develops serves a larger purpose. As he puts it, those who focus exclusively on statistics may “not work on interesting empirical problems,” while those who ignore theory risk “solving problems in an ad-hoc way.”
This holistic perspective distinguishes him from many of his peers, who often operate firmly within either the quantitative or qualitative camps. It also makes him a valuable bridge figure in an increasingly specialized academic environment—someone who can translate between the language of mathematics and the language of political science, between the demands of rigorous proof and the messier requirements of practical application.
His work has already earned him recognition: a broad portfolio of research, multiple awards, and a rapid rise through the academic ranks. But Egami seems less motivated by accolades than by the intrinsic satisfaction of solving hard problems. He describes the moment he first saw Imai present as a turning point, a moment when he realized that “even 20 percent” of that level of scholarship would be a dream. Having now achieved and exceeded that goal, he is focused on the questions that remain unanswered.
The Future of Measurement in an AI-Driven World
As artificial intelligence continues to infiltrate every aspect of social and economic life, the challenges Egami has identified will only become more pressing. AI systems are not just tools used to measure society; they are becoming active participants in it, shaping what people see, how they vote, and what they buy. This creates a feedback loop in which measurement tools influence the very phenomena they are designed to observe.
Egami’s framework offers a way to navigate this complexity. By insisting that researchers study AI tendencies with the same rigor they would apply to human subjects, he is building a discipline prepared for a future in which human and machine behavior are inextricably interwoven. His methodological innovations will likely become standard practice, much the way that robust standard errors and instrumental variables now sit at the core of applied economics and political science.
For those outside academia—in government, industry, and civil society—the practical takeaway is clear: accurate measurement of society requires more than sophisticated algorithms; it demands a disciplined, context-aware approach that treats every analytical tool, human or artificial, as a potential source of bias. The MIT researcher’s work is a masterclass in how to achieve that discipline.
His career also serves as a reminder that intellectual progress often comes from unexpected encounters and the willingness to pursue questions across disciplinary boundaries. A mistaken workshop in Tokyo, a conversation with a graduate student, a seminar that was never meant for him—these accidents became the foundation of a career dedicated to ensuring that when we attempt to quantify the social world, we do so with integrity. As the tools for measurement grow more powerful and more complex, that integrity will be the difference between seeing clearly and being deceived by the shadows on the wall.
