Audit • Profile • Align
We develop AI moral intelligence tools through Bayesian calibration methods to audit, profile, and align AI systems with human moral reasoning using Moral Foundations Theory.
Concern for harm and welfare of others
Justice, rights, and equal treatment
Group cohesion and in-group solidarity
Respect for hierarchy and tradition
Purity, dignity, and sacred values
According to Haidt's Moral Foundations Theory, moral reasoning across all cultures draws on the same universal dimensions — but in different proportions. No argument is purely one foundation.
Every act of argumentation — expressed in political discourse, company documents, product campaigns, or opinions shared on social media — encodes a characteristic mixture of these dimensions.
That mixture is the moral fingerprint. A crisis message spikes on Care and Loyalty. An ESG report foregrounds Fairness and Sanctity. Internal policy leans on Authority. Mapping that fingerprint — and tracking when it drifts — is what Moralytics makes measurable.
Skórski, M. · arXiv:2605.22660
Skórski, M. & Landowska, A. · UncertaiNLP @ EMNLP Workshops 2025
Skórski, M. & Landowska, A. · arXiv:2507.18523
Landowska, A., Budzynska, K., & Zhang, H. · Argumentation
Skórski, M., Landowska, A., & Rajda, K. · Humanities and Social Sciences Communications, 12(1)
Co-founded by Maciej Skórski and Alina Landowska
GitHub