About this project
In everyday life, our experiences unfold moment by moment—each shaped by patterns of thoughts, feelings, and motivations. These moments can be thought of as the basic building blocks of human experience, yet psychology has rarely studied them directly.
This project explores whether it’s possible to identify distinct, recurring types of moments—what we call archetypal states—that appear repeatedly within individuals and can also be meaningfully compared across people. Using data collected in real time from people’s daily lives (ecological momentary assessment, or EMA) and advanced computational tools, we aim to uncover the “hidden structure” of these momentary states.
Why does it matter?
Understanding moments matters because many psychological theories—from models of mental illness to approaches in psychotherapy—assume that discrete states play a key role in how symptoms emerge, persist, or change.
If we can reliably detect these states, we can:
Better describe mental health processes as they unfold in daily life.
Improve predictions about when someone is likely to experience distress.
Pinpoint the best moments for support, paving the way for just-in-time adaptive interventions (JITAIs) that respond exactly when they’re needed.
How it works
The project combines intensive daily-life data collection with cutting-edge analytic techniques:
Feature selection – Using EMA data, we identify a small but powerful set of features (e.g., emotions, motivations, appraisals) that capture a wide variety of subjective states.
Reliability testing – We run a two-wave EMA study to see whether the same archetypal states show up reliably both within the same person and across different people.
Clinical focus – We examine two disorders—social anxiety disorder and substance use disorder—that are likely to involve different sets of states, testing whether our approach can detect these differences.
We also incorporate passive sensing (e.g., movement, phone use) and explore methods for “borrowing” data across participants and datasets to improve classification.