Academic journal article Journal of the Association for Information Systems

PLS-Based Model Selection: The Role of Alternative Explanations in Information Systems Research

Academic journal article Journal of the Association for Information Systems

PLS-Based Model Selection: The Role of Alternative Explanations in Information Systems Research

Article excerpt

(ProQuest: ... denotes formulae omitted.)

1 Introduction

Sociotechnical systems within the purview of information systems (IS) research are inherently complex due to intricate underlying causal interactions and processes. Models are formal quantitative representations of theories and hypotheses that are devised to offer partial explanations of such complex systems (Lauenroth, 2003). Researchers build models that serve distinct goals in varied settings of interest to approximate these processes by abstracting away the details. This may result in the existence of multiple models reflecting varied theoretical lenses, levels of development, assumptions, interpretations, contexts, or even current fads. While this diversity enriches the literature, it also creates the threat of fragmentation of the field due to the individual researchers' "parental affections" for particular models (Chamberlin, 1890; Grover, 2013). The role of the scientific process, and model comparison in particular, is to overcome the human biases and sharpen our understanding over time by identifying and selecting the best approximating model(s) explaining a phenomenon, thereby bringing the field together (Purcell, 1992).

Philosophers of science have long realized the importance of considering alternative explanations (i.e., models) when researching certain phenomena. For example, Popper (1959) argued that considering alternative explanations (or "possible causes") is a crucial step prior to any attempt at the "falsification" of a theory. While pondering over the question of why some fields saw faster scientific advances than others, Platt (1964, p. 350) reasoned that,

"The conflict and exclusion of alternatives that is necessary to sharp inductive inference has been all too often a conflict between men, each with his single Ruling Theory. But whenever each man begins to have multiple working hypotheses, it becomes purely a conflict between ideas".

More recently, Nuzzo (2015) warned against cognitive fallacies that may lead researchers to make serious scientific errors, such as collecting evidence to support a specific hypothesis, not looking for evidence against it, and ignoring alternative explanations. To counter this, she calls researchers to explicitly consider plausible (i.e., motivated by theory) alternative explanations that are not just strawmen, but span models that offer theoretically justified alternatives for explaining the phenomenon under study.1

Alternative explanations can come in different forms and give rise to several models with different (or additional) antecedents and/or model relationships, all of which are plausible within the realm of the theoretical framework(s) under consideration. For example, researchers may derive alternative models from a single theory or multiple theories, such as Venkatesh, Morris, Davis, and Davis (2003) who relied on model comparison to benchmark the unified theory of acceptance and use of technology (UTAUT) model against alternatives. Similarly, Plouffe, Hulland, and Vandenbosch (2001) compared alternative models derived from the technology acceptance and the perceived characteristics of innovating frameworks to benchmark their explanatory power.

Alternative models may also emerge when considering theories in new contexts with unique variables and effects. In this vein, Johns (2006) and Alvesson and Kärreman (2007) note that new contexts can result in important changes in theories, such as rendering originally theorized relationships redundant or altering their magnitude, and/or creating new relationships by introducing new antecedents. For example, Venkatesh Thong, and Xu (2012) tailored the UTAUT to a consumer context by identifying additional constructs and relationships and compared their UTAUT2 model with the original model.

Finally, alternative models may be created when researchers seek to build conceptual bridges across related streams of inquiry to provide a holistic understanding of the phenomenon. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.