Behavioral Patterns and Social Media Explorations for Toxicology using Neurosymbolic AI Approaches

Abstract

Behavioral science is central to understanding how human behavioral patterns
emerge and are expressed within psychological contexts. A formal ontology of behavioral
patterns is therefore essential, as standardized vocabularies enable researchers to operate
within a consistent, interoperable framework.
Objectives: A general and comprehensive ontology for representing behavioral learning
and conditioning processes is still lacking. To address this gap, we introduce the Behavioral
Patterns Ontology (BPO), a foundational framework for formally representing psychological
concepts through patterns of behavior, with particular emphasis on the learning and conditioning
mechanisms that shape, reinforce, and modify behavior over time. Through this analysis we aim to
help identify trends, and assist toxicologists in better detecting unknown test results should new
substances arise.
Methods: The current version of BPO comprises 107 classes and 12 object properties, organized
through 106 hierarchical relationships and specified by 341 axioms. Together, these components
provide a structured representation of behavioral processes and their interdependencies. To
demonstrate its practical utility, we applied BPO to the annotation of TikTok video data and
integrated the resulting semantic annotations into a Social Media Analysis Dashboard for
visualization and interpretation at the individual-video level. Social media platform posts were
analyzed by domain experts and annotated by computer scientists with the help of our psychology
team. The data have been verified through domain expert knowledge. We further evaluated
the ontology using SPARQL queries driven by the ontology across multiple use cases and have
accuracy measures in place for each analysis.
Results: The results show that BPO effectively captures learning mechanisms, substancespecific
behavioral patterns, and the drivers and contextual settings associated with drug-seeking
behavior on TikTok. These findings suggest that BPO offers a promising foundation for the formal,
reusable, and computationally tractable representation of human behavioral patterns across digital
and psychological research settings.
Discussion: Although existing ontologies in the behavioral and social sciences provide
important domain-specific frameworks, including models of addiction, behavior change
interventions, mental functioning, emotions, and social entities, they are not designed to
systematically capture the cross-context mechanisms underlying behavioral formation and
adaptation. BPO addresses this unmet need by providing a transferable, semantically grounded
framework for representing behavioral processes across multiple application domains.