Understanding Adults' Spatial Cognitive Processes: A Time-Embedded N-Grams Model with Machine Learning

Area of Science:

• Educational Measurement
• Cognitive Science
• Data Science

Background:

• Technological advancements enable detailed collection and analysis of process data from educational assessments.
• Process data, capturing human-machine interactions, offers insights into cognitive processes during assessments.

Purpose of the Study:

• To explore factors influencing success/failure on spatial numeracy items.
• To identify spatial cognitive process features associated with different numeracy performance levels.
• To introduce a novel time-embedded n-grams model for analyzing sequential process data.

Main Methods:

• Utilized sequential process data from the 2012 Program for the International Assessment of Adult Competencies (PIAAC).
• Employed a time-embedded n-grams model, random forest, and XGBoost machine learning methods.
• Analyzed data from 596 U.S. adult respondents interacting with a spatial numeracy item.

Main Results:

• Time-related features and understanding directions in spatial dimensions were most predictive of numeracy skills.
• Machine learning models successfully predicted respondents' numeracy performance.
• Identified robust classifiers for distinguishing high and low numeracy performance levels.

Conclusions:

• Process data holds significant potential for analyzing latent numeracy skills and cognitive processes.
• The time-embedded n-grams model effectively incorporates temporal information into sequential action analysis.
• Findings underscore the importance of temporal and spatial cognitive factors in adult numeracy.