Framework & Rigor
Scientific Foundations & Methodology
Tiresias AI operates at the intersection of psychometric science, agent-based modeling, and demographic calibration. Here is the academic theory and empirical evidence behind our technology.
1. Grounded in Established Psychometrics
Skepticism is healthy in AI. Unlike models built on arbitrary generative heuristics, the Tiresias T-Score is a proprietary operationalization of the OCEAN (Big Five) and HEXACO personality frameworks. These are the most heavily researched, cross-culturally validated, and replicated taxonomies in the history of psychology.
Rather than asserting absolute individual prediction (which is scientifically untenable), Tiresias relies on probabilistic, population-level preference structures. At individual levels, personality traits act as biases; at cohort or aggregate scale, these individual variances smooth out, revealing extremely stable and predictable behavioral trends.
2. Empirical Validation and Benchmarks
To move beyond internal testing, we continuously validate Tiresias agent-swarm models against public, independent, real-world data:
- COVID-19 Compliance Hindcast: In a retrospective validation study, Tiresias predicted county-level public health compliance and vaccine uptake rates during the pandemic, achieving an equal-county Pearson correlation of r = 0.86 (and population-weighted correlation of up to r = 0.93) against ground-truth CDC and New York Times outcomes. Read the full COVID-19 Hindcast study.
- Media & Catalog Preference Lifts: Across representative validation runs (N=6,000 U.S. participants), Tiresias predictions outperformed traditional collaborative filtering recommendation engines by +39% in film, +57% in television, and +193% in books in top-K precision.
3. Peer-Reviewed Academic Literature
Tiresias translates established academic research into production-ready software. Our core modeling processes are grounded in the following peer-reviewed papers:
“The Do Re Mi's of Everyday Life: The Structure and Personality Correlates of Music Preferences”
Rentfrow, P. J., & Gosling, S. D. (2003) · Journal of Personality and Social Psychology, 84(6), 1236
Methodological RelevanceEstablished the primary empirical mapping demonstrating that Big Five personality traits (OCEAN) explain a significant portion of music and media preferences.
“Private traits and attributes are predictable from digital records of human behavior”
Kosinski, M., Stillwell, D., & Graepel, T. (2013) · Proceedings of the National Academy of Sciences (PNAS), 110(15), 5802-5805
Methodological RelevanceShowed that digital footprints (e.g. text patterns, online actions) can be used to infer latent psychographic traits with accuracy comparable to traditional test questionnaires.
“Musical preferences predict personality: Evidence from active listening and Facebook Likes”
Nave, G., Minxha, J., Greenberg, D. M., Kosinski, M., Stillwell, D., & Rentfrow, P. J. (2018) · Psychological Science, 29(7), 1147-1158
Methodological RelevanceReplicated personality-preference associations under large-scale behavioral observations, proving transferability and robustness across platforms.
“The spread of behavior in an online social network”
Centola, D. (2010) · Science, 329(5996), 1194-1197
Methodological RelevanceProvided mathematical models for complex behavioral contagion in small-world networks, which grounds our Watts-Strogatz agent-swarm contagion architectures.
“Personality trait structure as a human universal”
McCrae, R. R., & Costa, P. T. (1997) · American Psychologist, 52(5), 509
Methodological RelevanceConfirmed the structural cross-cultural stability and invariance of the Big Five framework, proving its reliability across international demographics.
4. Methodological Rigor & Limitations
In compliance with scientific rigor, we outline the boundaries and constraints of our models:
- Probabilistic, Not Deterministic: Personality metrics explain variance in preferences but do not determine individual choices in isolation. Our model outputs indicate a higher or lower density of interest within populations rather than predicting specific actions for individual citizens.
- First-Party Consent & Privacy: We strictly avoid invasive personal tracking. Tiresias does not utilize third-party cookies or device graphs. Individual client simulations are run on anonymized, aggregate, or consented data inputs, ensuring full compliance with GDPR, CCPA, and Apple's App Tracking Transparency policies.
- Continuous Recalibration: Cultural shifts, regional anomalies, and macroeconomic changes alter how personality traits express themselves. Tiresias regularly recalibrates demographic priors using the latest US Census American Community Survey (ACS) updates.
Request Scientific Documentation
Our briefing team regularly provides detailed data schemas, mathematical formulations of our Bayesian updates, and full replication protocols to academic partners, enterprise due diligence teams, and researchers.
Schedule a Science Briefing