MU study reveals impact of homework on student achievement

Thursday, September 19, 2024 - 12:00

Researchers at Maynooth University’s Hamilton Institute and Department of Mathematics and Statistics have unveiled significant findings on the role of homework in student achievement.

The research, led by Prof Andrew Parnell, Nathan McJames and Prof Ann O’Shea, used a new AI model to analyse data from the Trends in International Mathematics and Science Study (TIMSS 2019).

Focusing on 4,118 Irish students in their second year of secondary school, the research assessed the impact of varying homework patterns on their mathematics and science performance.

The study titled, ‘Little and Often: Causal Inference Machine Learning Demonstrates the Benefits of Homework for Improving Achievement in Mathematics and Science’ was published in the journal Learning & Instruction.

It offers new evidence on how homework frequency and duration affect academic performance among Irish secondary school students.

Key findings:

Frequency over duration: The study highlights that the frequency of homework is more important than its duration. Daily homework assignments were found to be most effective for improving mathematics achievement, while science performance benefitted most from homework assigned three to four times a week.

Effectiveness of shorter assignments: Short-duration homework tasks, lasting up to 15 minutes, were shown to be just as effective as longer assignments. This suggests that regular, concise homework  can promote learning without overwhelming students with excessive work.

Equity in benefits: Contrary to previous research, this study found that all students, regardless of socioeconomic background, experienced similar benefits from homework, indicating equitable advantages across diverse student populations. The researchers advocate for homework policies prioritising regular, short-duration assignments to optimise student engagement and academic success without causing undue stress.

Nathan McJames, the lead author, commented: “Our study provides strong evidence that regular homework can significantly enhance student performance, especially when given ‘little and often’. By avoiding very long homework assignments, this also allows students to balance schoolwork with other important activities outside of school.”

Prof Andrew Parnell added: “Our use of advanced causal inference methods ensures the reliability of our findings. This research provides valuable insights that can guide evidence-based policy changes in education, ultimately benefitting students across the board.”

Published in the journal, Learning & Instruction, this study used a special type of AI model to understand how homework affects students’ performance in maths and science. Here’s how it works:

1. Learning from data: The model looks at a lot of information about students (over 4,000 Irish 14 year olds), to see how often they do homework and how well they do in school. It tries to find the hidden patterns to see if doing homework more often helps students get better grades.

2. More flexible: Unlike traditional methods that can only capture simple relationships in the data, our new AI approach can handle more complicated situations. It doesn’t just guess that more homework is always better—it checks for different possibilities, and it allows for different students to have a different reaction to different levels of homework

3. Finding the true effect: The AI helps us figure out if homework really makes a difference or if other things, like how rich a student’s family is or how experienced their teacher is, are playing a bigger role. It separates these effects to understand what’s truly happening. 

4. Multiple subjects at once: Instead of looking at just one subject, like maths, this model can study the effects on multiple subjects simultaneously. In our case we use both maths and science at the same time as they were both recorded in the study. This helps get a clearer picture of how homework works differently in different subjects, whilst also searching for commonalities between the two. 

This research was supported by Science Foundation Ireland under grant number 18/CRT/6049. Additional support for Andrew Parnell was provided by a Science Foundation Ireland Career Development Award and SFI Research Centre award.

The published version of the research is available Open Access in Learning and Instruction at: https://doi.org/10.1016/j.learninstruc.2024.101968.