Poster #51 - Homework Behaviors Predicting Academic Achievement: Evidence from Process Data Based on Smartpen Technology

• In Event: Poster Session 14

Abstract

Substantial studies have indicated the positive relation between homework and academic achievement, but previous research has been limited to using students’ self-reported homework activity rather than objective measures on real-time behaviors. The current study employed smartpen technology for tracking the details of students’ homework behaviors and aimed to 1) validate the framework of approaches to learning based on homework process data; and 2) investigate the prediction from homework behaviors to academic achievement three months later.

The current study included a school sample of 270 fifth graders (age=10-11y, 51.5% boys) from Huai-an, China. Homework behaviors were assessed with process data collected by smartpens (i.e., dot-matrix pens, Model 130-T41) from 12 assignments in March, 2022 (including 4 assignments for Chinese, Math, and English, respectively). When doing homework, students used the smartpens writing on paper printed with dot-matrix dots, and the infrared high-speed camera capture technology embedded within the smartpen synchronized, transferred, and converted the information of handwriting on the paper. The information captured by smartpens were computed into 7 measures to characterize students’ homework behaviors, including 1) three measures for grit-related behaviors: Difficult problems attempted, Difficult problems second try, and Impulsive completion; 2) two measures for time-management related behaviors: Time arrangement, and Time spending efficiency; and 3) two measures for attention-related behaviors: Skipping question, and Balanced focus.

Academic achievement was assessed with students’ final exam scores of the semester collected from three subjects (including Chinese, Math, and English) at the end of June, 2022. The final exam scores were standardized to normal curve equivalents, and the total of three subjects was used in the modeling and regression analyses.

Hierarchical confirmatory factor analyses were conducted to validate the framework of approaches to learning based on homework process data. Three nested models were tested: Model 1) a hierarchical model of multitraits (CFI=.73, TLI=.69, RMSEA=.14, AIC=1171.09, as shown in black in Figure 1); Model 2) a hierarchical model of multitrait-multimethod, which added the three latent homework-setting variables to Model 1 (CFI=.87, TLI=.83, RMSEA=.10, AIC=723.29, the adding part as shown in blue in Figure 1); Model 3) a modified hierarchical model of Model 2, which added the covariances between parts of residuals (CFI=.93, TLI=.91, RMSEA=.07, AIC=527.24). The Model 3 was the best-fitting model, which indicated the structure of approaches to learning based on homework process data.

To test the prediction from homework behaviors to academic achievement, a composite score of the three subject grades was finally added to the Model 3 (as shown in green in Figure 1). The multilevel structural equation modeling analysis indicated that approaches to learning based on the homework process data significantly predicted academic achievement three months later (p<0.01). Stepwise regression analysis was conducted to further examine specific homework behaviors’ contributions to academic achievement. Results as shown in Table 1 suggested that a) in Chinese homework: impulsive completion and time arrangement, b) in Math homework: impulsive completion and time arrangement, c) in English homework: impulsive completion, difficult problems second try, and time spending efficiency significantly predicted academic achievement three months later.

Authors

• Manjie Wang, 17 Education and Technology Group
  Presenting Author

• Juan Liu

• Baitao Chen

• Yan Li