Model‐Based Optimization of Learning Curves: Implications for Business and Government. (28th October 2015)

• Record Type: Journal Article
• Authors: Madni, Azad M.; Paulson, Courtney; Spraragen, Marc; Richey, Michael C.; Nance, Marcus L.; Vander Wel, Michael
• Is Part Of: INCOSE International Symposium. Volume 25(2015)Supplement 1
• Page Start: 1070
• Page End: 1084
• Publication Date: 2015-10-28
• Subjects: Systems engineering -- Congresses; Systems engineering -- Periodicals; 620.0011
• DOI: 10.1002/j.2334-5837.2015.00116.x
• Languages: English
• ISSNs: 2334-5837

Abstract

Abstract: Major corporations such as Boeing spend substantial amounts annually on workforce training and development. They typically have a fixed budget to allocate to different learning options (e.g. online learning, in‐person instruction) available at different geographic locations. Currently, these organizations do not have the means to quantify and optimize the impact of their investments on employee time‐to‐proficiency and organizational learning curves. This paper presents an integrated model‐based framework that combines parametric curve‐fitting, portfolio optimization, system dynamics, and agent‐based modeling for learning curve sensitivity analysis and optimization. The framework enables assessment of the sensitivity of learning curves to changes in budget and learner allocations to the available learning options. Learning curves optimization takes into account the "forgetting" behavior that results from disruptions such as interruptions in learning arising from temporary re‐assignment of learners to take care of other organizational needs. Limited pre‐test and post‐test data from a lecture‐based course on composites pilot offered by Boeing overseas was used for model calibration and learning curves optimization. The ability to optimize learning curves for a specified budget and learning options has wide applicability in production engineering, operations, and customer‐facing functions. Future data collection is expected to include data from intermediate tests or responses to questionnaires in addition to pre‐test and post‐test data to achieve a more accurate curve‐fit.