Concerns over foreign electoral interference have grown since the 2016 U.S. presidential election, yet public-facing intelligence assessments continue to rely on vague probabilistic language that limits clarity, consistency, and operational insight. This study introduces an exploratory AI-facilitated framework designed to systematically quantify the likelihood of foreign election interference across U.S. elections from 2016 to 2036. Drawing on declassified intelligence assessments from the ODNI, NIC, and CISA, corroborated by open-source intelligence (OSINT), we applied a three-phase natural language processing (NLP) protocol using OpenAI’s tools to extract, classify, and scale linguistic indicators of confidence. These were then mapped to probabilistic values based on Sherman Kent’s CIA estimative language and modeled using Monte Carlo simulations to account for uncertainty. Named Entity Recognition and sentiment analysis identified country-specific attribution patterns, while lexical scaling translated narrative judgments into quantifiable interference probabilities. Results revealed persistently high likelihoods of Russian interference, alongside growing probabilistic signals from China and Iran over time. A hierarchical linear model confirmed significant variation by election year and actor, and simulation-based forecasts suggest increasing probabilistic risk through 2036. This framework offers a replicable, data-driven model for transforming qualitative intelligence into structured probability distributions, providing analysts and policymakers with an evidence-based tool to track, compare, and forecast adversarial influence strategies with greater transparency and granularity.
May et al. (2026) studied this question.