Abstraction Techniques Enhance Human Understanding of Symbolic AI Explanations
Researchers from computer science investigate how formal abstractions—removal and clustering of irrelevant details—affect human reasoning with symbolic AI explanations. Using Answer Set Programming (ASP), they defined irrelevant details to be abstracted for simplified explanations. Cognitive experiments with participants classifying stimuli across domains showed that clustering details significantly improved understanding, while removal reduced cognitive effort. The study supports the hypothesis that abstraction enhances human-centered symbolic explanations, addressing the high extraneous cognitive load of raw logical traces.
Key facts
- The study uses Answer Set Programming (ASP) as a formal framework.
- Two abstraction techniques were tested: removal and clustering of irrelevant details.
- Cognitive experiments involved participants classifying stimuli across domains.
- Clustering details significantly improved participants' understanding.
- Removal of details significantly reduced cognitive effort.
- The research aims to reduce extraneous cognitive load in symbolic AI explanations.
- The paper is categorized under Computer Science > Artificial Intelligence.
- The submission history is from arXiv.
Entities
Institutions
- arXiv