Towards a Paradigm Shift: How Can Machine Learning Extend the Boundaries of Quantitative Management Scholarship?

Abstract

Management scholarship is beginning to grapple with the growing popularity of machine learning (ML) as an analytical tool. While quantitative research in our discipline remains heavily influenced by positivist thinking and statistical modelling underpinned by null hypothesis significance testing, ML is increasingly used to solve technical, computationally demanding problems. In this paper, we argue for a wider, more systematic adoption of the key tenets of ML in quantitative management scholarship, both in conjunction with and, where appropriate, as an alternative to canonical forms of statistical modelling. We discuss how ML can extend the boundaries of quantitative management scholarship, help management scholars to unpack complex phenomena, and improve the overall trustworthiness of quantitative research. The paper provides a representative review of the use of ML to date and uses a worked example to demonstrate the value of ML for management scholarship.

Testing service infusion in manufacturing through machine learning techniques: looking back and forward

“…Following Valizade et al. (2022) suggestion to apply ML techniques specific to the research questions posed, in their case identifying the best predictors for innovation outcomes, and Chou et al.…”

“…The evolution of ML literature has focused on developing new algorithms that maximise accuracy metrics on isolated datasets, yet addressing the real impact of these techniques remains an open issue (Wagstaff, 2012). Valizade et al. (2022) advocate for the incorporation of new analytical tools and advanced statistical modelling based on null hypothesis significance testing, emphasising the need for a more balanced methodological approach.…”

“…, 2021). For instance, Valizade et al. (2022) applied ML techniques to predict outcomes in a European innovation survey, comparing parametric approaches like logistic regression with a range of algorithms.…”

“…The evolution of ML literature has focused on developing new algorithms that maximise accuracy metrics on isolated datasets, yet addressing the real impact of these techniques remains an open issue (Wagstaff, 2012). Valizade et al (2022) advocate for the incorporation of new analytical tools and advanced statistical modelling based on null hypothesis significance testing, emphasising the need for a more balanced methodological approach. ML offers a balanced avenue for theory generation and testing research, which is particularly beneficial for scrutinising and potentially providing robust evidence against prevailing assumptions and expectations (Tonidandel et al, 2018).…”

“…Unlike traditional statistical models like regression or SEM, ML techniques excel in unravelling complex phenomena by deciphering nonmonotonous and non-linear effects between model variables (Kalnins, 2022;Leavitt et al, 2021). For instance, Valizade et al (2022) applied ML techniques to predict outcomes in a European innovation survey, comparing parametric approaches like logistic regression with a range of algorithms. ML outperformed logistic regression by reporting variable importance scores, identifying non-linear relationships and effectively addressing multicollinearity issues typically encountered in regression analyses.…”