Presented by
Dr. Kai Larsen
Associate Professor- Systems Division
Leeds School of Business
University of Colorado
Abstract
From the birth of psychometrics to present day, the “identity” of latent constructs has remained an inaccessible puzzle to social and behavioral scientists. In fact, the Jingle and Jangle fallacies were identified as early as 1904 and 1927, respectively, stating that identically named constructs may not measure the same latent construct and that differently named constructs may measure the same latent construct. This article addresses the two fallacies as the Construct Identity Fallacy, and proceeds from empirical findings that this fallacy exists within the Information Systems discipline. Success in addressing the Construct Identity Fallacy would increase transparency in the behavioral and social disciplines and is material to cumulative research traditions. Currently, only two modes enable construct identity evaluation; Multitrait-Multimodel (MTMM) and factor analysis. Both modes require collection of empirical data through surveys before analysis, rendering them impractical for analysis of more than 100 constructs. A quantitative approach to construct identity detection is required. Based on the design science paradigm, we therefore propose and evaluate three IT artifacts, represented through different natural language processing algorithms. We use a dataset consisting of all the relevant behavioral studies from a set of two top IS journals over a 20-year span to evaluate our approach in comparison to human expert decisions. After evaluating the three IT artifacts, the two lower-performing artifacts are combined with the third artifact, and one of these is found to perform the best and named the Construct Identity Detector (CID). The article concludes with a description of a research program based on this method and its implications for IS as well as other disciplines, arguing that the IS discipline is uniquely positioned to address the Construct Identity Fallacy.
When
15 May 2012
12:00pm – 1:00pm
Where
University Centre, BLD (06_02_10)
Bond University
Contact Information
Catherine Smith
Database and Research Support
Faculty of Business
Telephone: +61 7 5595 55721
Bond University | Gold Coast, Queensland, 4229, Australia
