Abstract:
As probabilistic machine learning systems are deployed increasingly widely, computational literacy will require better understanding of the fundamental distinctions between conventional deterministic models of computation, and probabilistic behaviours. This may be especially important in cases that seem superficially similar, such as text string generation via sequence prediction in LLMs. We introduce NodeToJoy, a visualisation that juxtaposes a Finite State Machine visualisation with a Markov Chain visualisation, both adapted to the exploration of large and complex datasets by interactive hierarchical semantic zooming (including a novel extension of Harel’s Statecharts to Markov Chains). Our paper reports classroom studies demonstrating that lessons including interaction with NodeToJoy significantly advance student understanding of these fundamental principles.
PPIG 2026 - 37th Annual Workshop
NodeToJoy: Understanding Probabilistic State through Interactive Visualisation