Abstract:

It is widely accepted that promoting teacher quality and their classroom performance is a key element to improve student achievement. However, the latest studies show that more complex approach should be used to support student learning. Our guiding assumption is that the enhancement of the student performance depends on:
1) student motivation to learn;
2) teachers capability to adjust instructions;
3) school administration to provide sustainable leadership to school community.

The essential element of such approach is to provide appropriate feedback for all stakeholders. Here we focus on feedback for teachers and school administrations.

Most often the source of credible feedback are different forms of data describing student performance, however the data driven decision making to implement interventions mostly is based on data literacy skills of school administration and teachers. Thus, the automated analysis and humanization of student performance data is crucial to support data driven decisions for improving classroom-learning quality in school. Our aim in this paper is to illustrate a novel model for processing student performance data to provide feedback for school leaders and teachers. The model developed a validated in this study in future will be used as a framework for development of automated feedback system for school.

To develop and validate the model we analysed student performance data from annual national level language literacy and numeracy assessment diagnostic tests for 3rd and 6th grade in Latvia for the last three years (2018-2020). To find and understand patterns in data, i.e., student levels of performance and test item difficulty, useful for feedback generation, we used IRT (Rasch models) for each test (e.g. 3rd grade numeracy). The model were validated in 10 primary and secondary schools in Latvia.

Our findings show that at least five key points should be included in model for future automated feedback system for school administration and teachers. Data on student performance in school level can be analysed comparing:
1) mean student performance in school to national sample;
2) student performance between classes;
3) student performance between different subjects;
4) student skills to solve various tasks where different cognitive effort is needed;
5) student knowledge to deal with various content elements of subject.

After validation of model, we suggest that automatic feedback for teachers could be automatically generated at least in four dimensions:
1) basic teaching skills;
2) teaching instructions;
3) efficiency and productivity in classroom;
4) support for students to provide personalized learning.

Similarly, we find that feedback for school administration could be provide in three dimensions:
1) support for teachers community;
2) recommendations for personalized professional development programs for teachers;
3) support for students with learning difficulties and development of personalized learning path.

Keywords:

Feedback, automated feedback system, student achievement, school improvement, data-driven decision making.