As artificial intelligence moves into hiring, education and health care, a central scientific question is becoming impossible to ignore: Can machines measure people accurately, fairly and transparently? At Rice University, psychological scientist Tianjun Sun is working at the intersection of psychometrics and artificial intelligence to help answer that question. Her research focuses on developing more reliable ways to measure human characteristics—including personality, motivation, interests, knowledge and cognitive ability—while ensuring that AI systems used for those measurements meet rigorous scientific standards.
Sun, an assistant professor of psychological sciences at Rice, has received the 2026 Academy of Management Research Methods Division Lawrence R. James Early Career Achievement Award. The honor recognizes early-career scholars whose work has made distinguished contributions to research methods, including methodological research, practice, education and service. For Sun, the award is especially significant because the division includes researchers whose work has shaped how organizational scientists design studies, evaluate evidence and analyze data.
“I was both surprised and honored,” Sun said. “The Research Methods Division includes many outstanding scholars whose work has shaped how organizational researchers conduct studies and analyze data, so receiving this recognition is incredibly meaningful.” The award highlights a field that is often less visible than headline-making discoveries but is essential to trustworthy science: measurement. Without dependable methods for translating complex human traits into data, even sophisticated statistical models and AI systems can produce misleading conclusions.
“Good science depends on good measurement,” Sun said. “Before we can answer important questions about people, organizations or society, we need valid ways of measuring the constructs we care about.” In psychological science, many of the most important variables cannot be observed directly. Researchers cannot see motivation or personality in the same way they can observe body temperature or blood pressure. Instead, they use structured interviews, questionnaires, behavioral tasks and performance data as indicators of underlying traits. Psychometrics provides the technical framework for determining whether those indicators are valid, reliable and appropriate for the people and contexts in which they are used.
A measurement can be reliable without being valid. For example, an assessment might produce highly consistent results while measuring the wrong characteristic or favoring one group over another. Researchers therefore examine several forms of evidence, including whether an assessment captures the intended construct, whether scores predict relevant outcomes and whether the interpretation of those scores remains appropriate across populations. Fairness is another critical concern. A system may appear statistically accurate overall while producing systematically different errors for people from different demographic or cultural groups. Those errors can have serious consequences when assessments influence job opportunities, educational access, medical support or public policy.
Sun’s recent work addresses these challenges through what she describes as “psychometric AI,” an emerging area that combines psychological measurement theory with artificial intelligence. Psychometrics has traditionally emphasized construct validity, reliability, fairness and theoretical interpretation. AI research, by contrast, has often prioritized prediction, scalability and automation. Bringing the disciplines together means asking not only whether an algorithm can make a prediction, but also what exactly it is measuring, how its conclusions should be interpreted and whether its performance is equitable across different groups.
One project from Sun’s research program developed and validated an AI chatbot designed to assess personality through natural conversation. Rather than relying exclusively on fixed questions and standardized response options, the system can interact with a person conversationally and use the exchange to estimate psychological characteristics. That flexibility could make assessments more engaging and accessible, but it also introduces technical challenges. Conversational systems must distinguish meaningful behavioral signals from irrelevant language patterns, account for differences in communication style and avoid treating fluency, cultural expression or familiarity with technology as direct evidence of personality.
Validation is therefore central to the project. An AI assessment cannot be considered scientifically credible simply because it produces plausible descriptions or agrees with human impressions. Researchers must determine whether its scores correspond to established psychological constructs, whether results are sufficiently stable, how much information the system requires and whether it performs consistently across populations. They must also examine the possibility that the system is relying on hidden proxies, such as accent, vocabulary, education or demographic cues, rather than the trait it claims to measure. These evaluations help establish whether conversational AI is genuinely measuring personality or merely recognizing patterns associated with particular groups.
Sun’s research also reaches into health care, where conversational AI could one day support the assessment of cognitive functioning and the detection of early signs of cognitive decline. A system that analyzes speech, responses, pauses or changes in conversational behavior might provide an accessible way to identify people who could benefit from further clinical evaluation. Such tools would not replace professional diagnosis, but they could help expand screening and make it easier to recognize potential problems earlier. At the same time, health-related applications require especially careful safeguards because errors, privacy violations and unexplained recommendations can affect patients’ treatment and well-being.
The broader promise of Sun’s work is to make AI-based assessment more scientifically grounded before these technologies become deeply embedded in everyday decisions. Her laboratory is exploring AI-assisted interviews, adaptive conversational assessments, educational technologies, cognitive health assessment and methods for making AI systems safer and more interpretable. Adaptive systems could change the next question based on a person’s previous response, potentially gathering useful information more efficiently than a fixed test. Yet greater flexibility also increases the need for clear documentation, reproducible methods and human oversight so that users understand how conclusions are generated.
As AI systems become more powerful, Sun argues that innovation should not come at the expense of measurement principles developed over decades of psychological research. “We are working to make sure that AI-based assessments are scientifically valid, trustworthy and fair,” she said. “If these technologies are going to influence important decisions about people’s lives, they need to meet the same rigorous standards that we expect from traditional methods.” Rice’s collaborative environment, including connections across psychology, computer science, engineering and data science, has helped support that effort. The outcome could be a new generation of AI tools that do more than predict human behavior: They could measure it with greater accuracy, explainability and fairness.
Subject of Research: Psychometric artificial intelligence; AI-based assessment of personality, cognitive functioning and human abilities; measurement validity, reliability, fairness and interpretability.
Article Title: Rice Researcher Develops Scientific Standards for AI That Measures Human Behavior
Web References: https://profiles.rice.edu/faculty/tianjun-sun; https://str.aom.org/discussion/2026-aom-research-methods-division-award-nominations-5; https://www.aom.org/
References: Academy of Management Research Methods Division Lawrence R. James Early Career Achievement Award; Rice University profile of Tianjun Sun.
Keywords: Artificial intelligence, generative AI, machine learning, psychometrics, psychological assessment, personality measurement, cognitive health, fairness, AI ethics, Rice University, Tianjun Sun
