The Four Pillars of Scientific Discovery and Their Role in Undergraduate Data Science Curricula
Ivaylo Donchev *
Faculty of Mathematics and Informatics, “St. Cyril and St. Methodius” University of Veliko Tarnovo, Veliko Tarnovo, Bulgaria.
*Author to whom correspondence should be addressed.
Abstract
The philosophy and practice of science have undergone profound methodological shifts over centuries. Although often viewed as a linear progression, the evolution of scientific discovery is better understood as an accretion of distinct paradigms: empirical, theoretical, computational, and data-driven. Each paradigm has unique epistemic foundations, tools, strengths, and limitations. The empirical paradigm grounds knowledge in direct observation and experimentation. The theoretical paradigm abstracts nature into mathematical laws and first principles. The computational paradigm uses simulation to explore systems that are intractable through pure theory alone. The data-driven paradigm extracts patterns from large datasets using machine learning, often without a priori hypotheses. This paper is a conceptual framework analysis rather than an empirical curriculum evaluation. It argues that contemporary scientific progress increasingly depends on the fluid integration of all four paradigms rather than on competition among them. We characterise each paradigm, illustrate their interplay through concrete examples (drug discovery and climate science), and extend the analysis to undergraduate data science education. Drawing on the Computing Competencies for Undergraduate Data Science Curricula report (ACM, 2021) as a primary curricular reference, we map each paradigm to specific knowledge areas and competencies. The mapping is illustrative rather than exhaustive; other frameworks (e.g., statistical education guidelines) may complement this view. We conclude that paradigm integration is not merely a methodological ideal but a core curricular requirement. To support implementation, we recommend: (1) explicit paradigm mapping of courses and learning outcomes; (2) integrative capstone projects that require the application of all four paradigms; (3) cross-paradigm assessment that evaluates not only technical execution but also justification of paradigm choice; and (4) embedding ethical responsibilities—bias, opacity, assumptions, and privacy—as intrinsic to each paradigm rather than as an add-on. Graduates who can navigate across all four paradigms will be better prepared for the interdisciplinary, ethical, and scalable demands of modern data science.
Keywords: Scientific paradigms, empirical method, theoretical science, computational simulation, data-driven discovery, data science education, undergraduate curriculum, ACM competencies, curriculum design, science education.