The Use of Algorithmic Models to Develop Secondary Teachers’ Understanding of the Statistical Modeling Process

Abstract

Statistical modeling continues to gain prominence in the secondary curriculum, and recent recommendations to emphasize data science and computational thinking may soon position algorithmic models into the school curriculum. Many teachers' preparation for and experiences teaching statistical modeling have focused on probabilistic models. Subsequently, much of the research literature related to teachers' understanding has focused on probabilistic models. This study explores the extent to which secondary statistics teachers appear to understand ideas of statistical modeling, specifically the processes of model building and evaluation, when introduced using classification trees, a type of algorithmic model. Results of this study suggest that while teachers were able to read and build classification tree models, they experienced more difficulty when evaluating models. Further research could continue to explore possible learning trajectories, technology tools, and pedagogical approaches for using classification trees to introduce ideas of statistical modeling.

Introducing students to machine learning with decision trees using CODAP and Jupyter Notebooks

“…As Sulmont et al [20] found for non‐majors at the university level, it is more difficult to teach reasoning about machine learning models than it is to teach the algorithms, yet it is possible [19]. Regarding teacher education, Zieffler et al [22] recently noted that, analogously for mathematics teachers, decision tree algorithms can be taught well in in‐service training, but the evaluation of decision trees is a greater challenge. In our module, we focus on both decision tree algorithms and their evaluation.…”

Teaching and Learning Data-Driven Machine Learning With Educationally Designed Jupyter Notebooks

“…Using machine learning for predictive modeling requires two things: the technical ability to create a predictive model and the ability to reason about the process to decide whether the resulting model is useful or just wrong, to refer again to the aphorism from the beginning of this paper. Different studies similarly report that difficulties often occur with reasoning about machine learning processes (Sulmont et al, 2019b;Zieffler et al, 2021) and with documentation in computational notebooks (Rule et al, 2018). Our findings suggest that our approach consisting of three aspects (teaching module, study a worked example, use the worked example as a scaffold) to address these difficulties seems to be promising to support students in applying, documenting, and reasoning about a machine learning process; here exemplified by the machine learning method of decision trees.…”

“…We intend to study how our teaching module (including our ProDaBi Decision Tree JNBs) and our worked example support students. Since Sulmont et al (2019a) and Zieffler et al (2021) found that it is challenging for students to reason about machine learning models, we are interested in how a worked example supports upper secondary students in creating, evaluating, and reasoning about a decision tree model after attending our teaching module. We wonder if they can use an interactive worked example to transfer a machine learning modeling process to a new problem and reason about it adequately.…”

“…Sulmont et al, (2019a, p. 953) found, for non-major students at the university level, that "it is possible to teach ML to those with little to no math/CS background" but that they are "having difficulty appreciating the human decision-making aspects of ML." Sulmont et al (2019b) note that barriers for students are not related to understanding machine learning algorithms but that they get "stuck on the development and evaluation of models" (p. 14) Regarding teacher education, Zieffler et al (2021) found that decision tree algorithms as such can be well taught in in-service teacher education, but the evaluation of decision trees poses a more significant challenge. We designed a course for upper secondary students to enable them to use machine learning for predictive modeling, e.g., in the form of automatically created decision trees, for modeling and to enable them to reason about the outcomes.…”

Introducing High School Statistics Teachers to Predictive Modelling by Exploring Dynamic Movie Ratings Data: A Focus on Task Design

“…Education researchers are re-thinking and expanding their ideas about data and approaches to statistical modelling and have suggested the inclusion of machine learning approaches and associated algorithmic models in high school curricula (e.g., Biehler & Schulte, 2017). From a learning perspective, algorithmic models could offer a more accessible and conceptually simpler mechanism to introduce students to data science than inferential methods (Gould, 2017;Ridgway, 2016), and research by Zieffler et al (2021) suggested there might be similar benefits for high school statistics teachers. Predictive modelling, with its focus on developing models by learning from features of data to make predictions and forecasts for likely future outcomes, could also provide opportunities for students to integrate both computational and statistical thinking (e.g., De Veaux et al, 2017).…”