Integrating Interdisciplinarity in Higher Education: Lessons from Machine Learning Education
Introduction
Bridging Technical and Non-Technical Fields in AI Education
As the convenor of the master’s level module titled “Interdisciplinary Approaches to Machine Learning” at the Centre for Interdisciplinary Methodologies (CIM), Michael confronts the challenge of making technical knowledge accessible to non-technical students while keeping it engaging for those with technical expertise. Drawing from his diverse background in computer science, sociology, history of technology studies, computing developments in finance, and media theory, Michael aims to empower students to critically assess real-world claims about Artificial Intelligence (AI).
Inspired in part by a course taught by data scientists Jeremy Howard and Rachel Thomas at the University of San Francisco, Michael seeks to offer a more inclusive approach than the one he experienced, which he found exclusive to only those deeply embedded in computer science culture. His goal is to make the material even more accessible, particularly for students from diverse undergraduate backgrounds at CIM.
The module focuses primarily on recent developments in deep learning, commonly referred to as AI, teaching fundamental deep learning techniques while exploring their connection to social theory. Discussions centre on the social impacts and ethical considerations of AI applications, such as privacy, bias, accountability, and the influence of tech companies such as Google and Meta (previously Facebook). Michael emphasises that this isn't a typical computer science class on machine learning but rather uses deep learning as a starting point to examine how technical aspects of AI intersect with public concerns.
Dr. Michael Castelle
Principles of Practice
Facilitating Interdisciplinary Learning through Reading Sessions
Michael has found that the “strategy of reading sessions,” which creates a “safe space for ideas from different disciplines,” is highly effective for interdisciplinary teaching. This pedagogical approach emerged from the integration of technology for interdisciplinary learning and teaching during the pandemic. Initially implemented to foster social engagement by assigning students to random small reading groups online, this strategy yielded unexpected academic benefits. It encouraged students to engage in meaningful discussions about the reading material, leading to increased exploration of topics such as the social impact of AI within Michael's module.
Varied Assessments and Student Choice in Module Evaluation
Within the module, students encounter a diverse array of assessments. These range from group presentations focused on specific topics to mid-term assignments assessing their comprehension of module content. Culminating in the final project or paper, where students have the autonomy to select their own topic, this approach can pose challenges in accessibility, particularly in assessment moderation. Ensuring fairness and consistency in grading can be “slightly challenging” due to the subjective nature of evaluating diverse topics. However, throughout the process, students receive comprehensive guidance and support.
Michael's experience highlights that students often gravitate towards projects “leaning to their backgrounds”. Yet, he notes that some students do venture into new directions by integrating knowledge from other disciplines.
Moving Forward
Integrating Interdisciplinarity into Research and Teaching Practices
Michael emphasises that interdisciplinarity is “not something that you can just inject into teaching” but rather it must be integrated into one’s research. He encourages making interdisciplinarity a core aspect of research practice. This perspective stems from his experience co-editing the book “The Cultural Life of Machine Learning: An Incursion into Critical AI Studies.” This book brings together historians, sociologists, and scholars from media studies, communication studies, cultural studies, and information studies to explore the origins, practices, and potential futures of contemporary machine learning.
Michael encourages embracing “other perspectives” to cultivate proficiency across multiple disciplines and advises educators to “not be afraid of a heterogeneous syllabus.” He acknowledges that students may initially find this approach unfamiliar, but stresses that normalising it can relieve their concerns. This is particularly relevant for students who feel they are not ready for interdisciplinary classes because it does not align with their specific career goals. Additionally, he recommends engaging with students to understand their backgrounds, which facilitates effective teaching tailored to their needs. To enhance accessibility, he suggests employing varied explanations and terminology familiar to students from diverse backgrounds.
Advocating Clarity and Engagement in Interdisciplinary Teaching
Michael stresses the criticality for tutors to keep students “on board” with their teaching, ensuring clarity and engagement when explaining concepts from diverse backgrounds. When communicating with a non-interdisciplinary audience, he advocates convincing them of the importance of other fields and diverse perspectives on issues. Despite their perception of limited exposure to interdisciplinary learning, Michael highlights that such exposure is inevitable. He advises transparency in acknowledging the current scenario where interdisciplinary approaches may not be fully embraced. Importantly, he emphasises that promoting interdisciplinary perspectives enhances rather than diminishes the significance of disciplinary learning by integrating diverse viewpoints effectively.
