XAlgo: a Design Probe of Explaining Algorithms' Internal States via Question-Answering
summary
XAlgo models deterministic algorithms by parsing questions into five types, mapping them to a pseudocode-derived hierarchical directive a cyclic graph, and generating answers; a Wizard-of-Oz study with 18 participants (9 simulated XAlgo, 9 control) yields 88 questions total: 37 concept queries and 51 state queries (21 causality/rationale, 17 confirmation, 12 description, 1 contrast). Users struggle to formulate questions, find responses too wordy, and show no learning gains in multiple-choice or walkthrough scores, suggesting need for proactive, concise support.
AI-written summary · about