Validation Testing of Autonomous Learning Systems
Paul Collopy, Valerie B. Sitterle, Jennifer Petrillo
Abstract
ABSTRACT Validating artificial intelligence systems is challenging. Systems continuing to learn new behaviors in the field are particularly difficult, because the designer does not know what the functions need proving. This article reports on a research project using the Toulmin augmentation method to structure validation test plans and validation testing for very difficult systems like field‐learning AI.
Topics & Concepts
Field (mathematics)Computer scienceTest (biology)Artificial intelligenceSystems engineeringSoftware engineeringEngineeringMathematicsPaleontologyPure mathematicsBiologySoftware Reliability and Analysis ResearchSoftware Testing and Debugging TechniquesSafety Systems Engineering in Autonomy