Modeling the Macro-Behavior of Learning Object Repositories

Abstract

Introduction The publication of learning materials in online repositories is usually regarded as a simple process. To publish, the contributor provides or uploads the material (or the reference to the material), fills some metadata about the material, and then the material is available in the repository for others to find and reuse. The contributor can repeat this process for more materials as desired, while he or she is still interested in providing content to the repository. These seemingly simple processes that determine the micro-behavior of contributors and consumers give rise to complex macro-behavior at the repository level once the contribution and preference of hundreds or thousands of individuals is aggregated (Ochoa & Duval, 2008). For example, some learning object repositories grow linearly while others, having a similar number of contributors, grow exponentially. Also, the number of objects published by a given contributor is distributed differently depending on the kind of repository, but always following a long-tailed distribution (Anderson, 2006). Unfortunately, there is no research available about how the micro-behavior of the individuals is related to the observed macro-behavior of Learning Object Repositories. The fields of Bibliometrics and Scientometrics have been studying a similar problem: the process of paper publication in different venues (journals, conferences, repositories, etc.). In these fields, several models have been proposed to attempt to explain the observed patterns in the data. For example, De Price Sola (1976) proposed "Cumulative advantage" as a model to explain the inverse-power law distribution, also called Lotka by Coile (1977), observed in the number of papers published by a scientist in a given field. Egghe and Rousseau (1995) and Egghe (2005) refine this notion with the "success breeds success" model. However, the models used for scientific publication cannot be transferred to learning object publication because one of their main characteristics, the increasing rate of production observed in most successful scientific contributors, has not been observed in learning material contributors elsewhere (Ochoa & Duval, 2008). Nonetheless, the methodologies to establish and validate these models will be borrowed and re-used in the present study. The present work proposes an initial model to explain the macro-behavior of LORs based on the characteristics of their contributor base. This paper is structured as follows: the modeling section presents previous unexplained characteristics of Learning Object Repositories that this work proposes to model. In the next section the model is formally defined and explained. The validation section studies the model, comparing its predictions against empirical data. The paper ends with a discussion of the relevance of this model and further research needed to improve it. Modeling the Publication Process In a previous work (Ochoa & Duval, 2008), several characteristics of the publication of learning objects were measured. That work used data collected from several sources: * three Learning Object Repositories (LORp): Ariadne, Connexions and Maricopa Exchange; * three Learning Object Referatories (LORf): Merlot, Intute and Ferl First; * two Open Courseware sties (OCW): MIT OCW and OpenLearn and * one Learning Management System (LMS): SIDWeb. The findings of that work could be summarized as: * LORp and LORf grow in number of objects linearly in two stages (bi-phase linearly), but OCW and LMS grow exponentially. * Most LORp and LORf grow bi-phase linearly in the number of contributors. OCW and LMS grow exponentially. * The number of objects published by a given author follows a Lotka distribution with exponential decay in the case of LORp and LORf. OCW and LMS present a Weibull distribution. * The rate at which contributors publish materials followed a Log-Normal distribution for all the repositories studied. …