BackHome
Refresh: Off
Paper Summary
Share...

Direct link:

Integrating Artificial Intelligence/Machine Learning Into History Classrooms: Reasoning About Data Bias Through Modeling With Primary Sources (Poster 3)

Fri, April 12, 9:35 to 11:05am, Pennsylvania Convention Center, Floor: Level 100, Room 115B

Abstract

Objectives: Incorporating AIML (artificial intelligence/machine learning) activities into the K-12 curriculum can help prepare students for careers in AIML-related fields and promote the development of problem-solving and computational skills (Lin & Van Brummelen, 2021). While AIML is often associated with STEM fields, its applications are widespread and can be integrated into non-STEM subjects as well (Gresse von Wangenheim et al., 2021). Contributing to this line of research, this study explored the integration of AI into history classrooms, focusing on engaging students in exploring and building machine learning models using primary sources as text data.
Theoretical framework: We draw on the theoretical lens of inquiry-based learning to examine students’ inquiry processes. This theoretical lens allows us to understand that learning is a process of active engagement and construction of meaning (De Jong & Van Joolingen, 1998). The constructivist perspective recognizes that students come to the learning environment with prior knowledge and experiences, which shape their understanding of new concepts (Avsec & Kocijancic, 2016).
Methods and Data Sources: This study was conducted in an urban public high school in the Southeastern United States. In this study, 41 students used StoryQ—a free web-based machine learning and text mining technology for K-12 students (Jiang et al., 2023)—to explore the historical topic of redlining, using the redlining dataset. The term of redlining originated in the 1930s, when the federal government created maps to designate neighborhoods considered too risky for investment (D neighborhoods). We collected multiple data sources, including pre- and post-surveys, semi-structured interviews, field notes, and student-generated work. We utilized Braun and Clarke’s 6-step approach (Braun & Clarke, 2006) to conduct thematic analysis on these data sources.
Results: Our analysis indicates that integrating AIML into history classrooms could potentially engage students in critical inquiry and develop their analytical skills, in particular skills in recognizing biases in data sources and understanding the importance of context and language nuances in historical analysis and AIML technology development. Specifically, students shared that they were aware of the perspectives that the data represented and emphasized the importance of having a variety of perspectives in data sources. Understanding whose perspective data represents is crucial for both developing historical thinking skills and AI understanding. In historical thinking, it is essential to examine primary sources and understand the context in which they were created to interpret events and phenomena accurately. Similarly, in AI, understanding the perspectives represented by the data used to train models is critical in ensuring that the models are unbiased and do not perpetuate discrimination or reinforce harmful stereotypes.
Significance: Contributing to the interdisciplinary learning of AIML (e.g., Ruppert et al., 2023), this study highlights the importance of incorporating a historical perspective in understanding data bias in AI systems (e.g., how it is generated, for what purposes, and whose perspective is being represented). This could help students gain a nuanced understanding of the ethical and social implications of AIML, enabling them to critically evaluate the use of AIML in various contexts and make informed decisions about its applications.

Authors