Strategy Instruction in Early Childhood Math Software: Detecting and Teaching Single-digit Addition Strategies

• Main Author: Carpenter, Kara Kilmartin
• Language: English
• Published: 2013
• Subjects: Cognitive psychology; Educational technology; Mathematics > Study and teaching
• Online Access: https://doi.org/10.7916/D8FR03TN

In early childhood mathematics, strategy-use is an important indicator of children's conceptual understanding and is a strong predictor of later math performance. Strategy instruction is common in many national curricula, yet is virtually absent from most math software. The current study describes the design of one software activity teaching single-digit addition strategies. The study explores the effectiveness of the software in detecting the strategies first-graders use and teaching them to use more efficient strategies. Instead of a business-as-usual control group, the study explores the effects of one aspect of the software: the pedagogical agent, investigating whether multiple agents are more effective than a single agent when teaching about multiple strategies. The study finds that while children do not accurately report their own strategies, the software log is able to detect the strategies that children use and is particularly adept at detecting the effective use of an advanced strategy with a model that performs 67% better than chance. Overall, children improve in their accuracy, speed, and use of advanced strategies. Of the three teaching tools available to the children, the count on tool was most effective in encouraging use of an advanced strategy, highlighting a need to revise the other tools. Low-performers correctly used advanced strategies more frequently across the six sessions, while mid-performers improved after just one session and high-performers' correct use of an advanced strategy was consistent across the sessions. Whether a student saw lessons featuring a single agent or multiple agents did not have strong effects on performance. More research is needed to improve the strategy detection models, refine the tools and lessons, and explore other features of the software.