VotestratesML: Social Studies as a Vehicle for Teaching Machine Learning

Research output: Contribution to book/anthology/report/proceedingArticle in proceedingsResearchpeer-review

In this demonstration, we present VotestratesML; a collaborative learning tool for teaching machine learning (ML) in a high school Social Studies classroom using voter profile data. As ML becomes increasingly widespread in society (O’Neil, 2017), the importance of ML-literacy increases, and while traditional Computational Thinking (CT) as popularised by Wing (2006) covers technical aspects, a focus is needed on how ML changes society and our lives. We seek to Computationally Empower (Dindler, Smith & Iversen, 2020; Iversen, Smith, & Dindler, 2018) students to take part in the technological development and engage them in the larger questions about ML’s role in democratic societies. Based on the goal of Scandinavian high schools to prepare and empower students to participate in democratic society, we investigate how the Social Studies classroom can be used as a vehicle to support students’ learning and critical reflection about ML. We have designed VotestratesML; a construction kit for ML, where Social Studies students use a web-application to explore the role of ML in political campaigns by constructing their own models. The design of VotestratesML is based on deconstructionism (J. M. Griffin, 2018) and specifically J. Griffin, Kaplan, and Burke (2012)’s explore’ems; interfaces designed to let students and teachers explore a technology, even if unfamiliar with its core concepts. In VotestratesML, students collaborate in small groups to create the best possible model for predicting voter behaviour by tinkering with the data-set, features and model parameters to explore how different aspects of a ML model influence its predictions. Students are encouraged to draw on their Social Studies expertise, such as considering existing theories of voter-behaviour when selecting features in the real-world data-set. Following the group work, students compare their models by predicting the voter behaviour of a predetermined set of voter-personas and discuss the implications of using such models in political campaigns. We have deployed VotestratesML in a study involving two Danish high school Social Studies classrooms and a total of 61 students, aged 17-20, in a three-lecture unit on the use of ML in political campaigns. We found that the contextualisation of VotestratesML as a Social Studies-specific tool was successful in motivating students in working with ML. We also found that students were able to use their Social Studies vocabulary to argue for choices made while working with VotestratesML, that they formed more nuanced views on the use of ML and were able to reflect on the pros and cons of using machine learning in political campaigns. During the demonstration of VotestratesML, attendees will gain hands-on experience with VotestratesML, build their own ML-models for predicting voter behaviour, and be invited to participate in our on-going discussions on Computationally Empowering students in reflecting on the role of emerging technologies and in particular machine learning in society and their everyday lives.
Original languageDanish
Title of host publicationProceedings of the 2020 Constructionism Conference
Number of pages2
Publication year2020
Pages44-45
ISBN (Electronic)978-1-911566-09-0
Publication statusPublished - 2020

See relations at Aarhus University Citationformats

ID: 190375255