Opacity, obscurity, and the geometry of question-asking

Area of Science:

• Cognitive Science
• Computational Linguistics
• Information Retrieval

Background:

• Question asking is fundamental to human interaction and learning.
• However, the factors contributing to question difficulty remain poorly understood.
• Existing models often focus on prior expectations, potentially overlooking other crucial elements.

Purpose of the Study:

• To identify and quantify the dimensions of question difficulty.
• To explore the interplay between answer rarity and cue indirectness in determining difficulty.
• To propose a more comprehensive framework for understanding question-answering processes.

Main Methods:

• Analysis of two large datasets: New York Times crosswords and Jeopardy quiz questions.
• Operationalization of cue indirectness using word2vec embeddings.
• Regression analysis to identify multiple dimensions of question difficulty.

Main Results:

• Two primary dimensions of question difficulty were identified: obscurity (answer rarity) and opacity (cue indirectness).
• Opacity plays a significant role, highlighting the importance of synergistic information in resolving ambiguity.
• Additional factors include question complexity, answer network density, cue intersection, and signal words.

Conclusions:

• Question difficulty is multi-dimensional, extending beyond prior expectations.
• Effective question design can leverage contextual cues to aid comprehension.
• Understanding these dimensions can enhance Bayesian models and inform heuristics for navigating conceptual spaces.