Deep dives on psycholinguistics, NLP fraud detection, and the signals hidden in how people write when they lie.
SAR narratives are the primary evidence trail in financial crime investigations — and where sophisticated actors craft their most careful language. Current AML keyword screening catches structuring and layering but misses narrative manipulation. Here's how BSA officers and compliance teams use psycholinguistic analysis to surface deception that rule-based systems miss entirely.
Keyword filters are reactive. Sophisticated bad actors write clean — no prohibited terms, no obvious signals — and pass through every rule you've written. Psycholinguistic analysis detects the manipulation structure in how they write, not just what they write. A technical guide for T&S teams at marketplaces, fintech, and social platforms.
Insurance fraud costs $308B/year. SIU teams can't read every claim — but fraudulent narratives produce measurable linguistic patterns that human reviewers miss at scale. Here's how Candor's API flags high-risk claims, with an example analysis and integration walkthrough.
Anyone can build an AI that outputs a score. Proving it works is harder. We benchmark Candor against the LIAR dataset — 8,041 expert-labeled political statements from Politifact — and report exactly what F1=0.534 means, what it doesn't, and how we got there.
Traditional fraud detection misses the linguistic tells that cognitive science has been cataloguing for decades. We break down six signal categories — pronoun distancing, hedging, emotional leakage, cognitive complexity, detail specificity, and negation patterns — and explain how they power a production deception detection API.