Using Bayesian Nominal Indicator Modeling to Analyze Factors of Expansive Learning Environment in the Context of Finnish Apprenticeship Education

• In Event: 58.082 - Roundtable Session 25
  In Roundtable Session: 58.082-9 - Methodological Advancements in Quantitative Research on Work-Related Learning and Professional Development

Abstract

The study examines the role of environmental conditions in the development of vocational talent by focusing on the meanings that are given to work environment. Eraut (2000, 2004, 2007) and Fuller and Unwin (2003, 2011a, 2011b) have identified a number of factors that promote learning in the workplace. Fuller and Unwin’s research found that an expansive work environment, as opposed to a restrictive work environment, is one that is characterized by a number of features that will create more, stronger, and richer learning opportunities for a worker to develop a greater breadth and depth of knowledge and skills.

An online “Workplace as Learning Environment“ (WLE) survey (James & Holmes, 2012) contains 22 items on a 5-point response scale, identifying seven factors that constitute an expansive working environment: 1) Participation and understanding of the workplace; 2) Task performance; 3) Access to resources to help learning; 4) Judgment, decision-making, problem-solving and reflection; 5) Experience, task transition and career progression; 6) Status as a worker and a learner; and 7) Organisational development. WLE was submitted to all Finnish secondary level vocational institutions in 2014. Total number of respondents was 1559, but this paper focuses on 305 apprentice education students who completed the survey (233 females, 76.4%, 72 males, 23.6%; average age 40.4 years, SD=10.431).

We applied a data mining approach, namely Bayesian Dependency Modeling (Myllymäki, Silander, Tirri & Uronen, 2002), to investigate if the seven WLE dimensions exist in the empirical data, and if they do, are the dimensions related to each other. Results of the algorithmic data modeling showed that the most probable Bayesian Network (BN) contained all the seven dimensions. Investigation of the BN’s structure showed that versatile and challenging work tasks in collaboration with (and receiving feedback from) other workers are positively related to experiences of getting various forms of support (mentoring, training, material related to work tasks, encouragement to gain qualifications) in the workplace.

Relevance of apprenticeship education was investigated with Bayesian Classification Modeling. The class variable in the analysis was a single item in the survey (“Apprenticeship education taught me necessary working life skills”) with three categories (1=Disagree; 2=Neutral; 3=Agree). First phase of the analysis was to examine with genetic algorithm approach if the WLE dimensions are useful predictors of the class variable. Resulting BN (classification accuracy: 64.6%) contained three out of seven WLE dimensions, showing that apprentices who are satisfied with their on-the-job training self-assessed significant improvement in skills related to judgment, decision-making, problem solving and reflection. Further, they experienced that they had gained valuable work related experience and were treated in their workplaces both as workers and learners.

We conclude firstly that the model by Fuller and Unwin (2003) is generic and works in the Finnish context, and secondly that Bayesian discrete methods as an example of “algorithmic modeling culture” (Breiman, 2001) allow robust and meaningful analysis of categorical data (Nokelainen & Ruohotie, 2009).

Authors

• Petri J. Nokelainen, Tampere University of Technology

• Laura Pylväs, University of Tampere