An interference-adjusted power learning curve for tasks with cognitive and motor elements

Funding Information: The first author thanks the Finnish Work Environment Fund (No. 200224) for supporting this research. The second author sincerely thanks Prof. Charles D. Bailey of James Madison University for making his data available. Without it, this study would not have been possible. He also conveys his thanks to the Social Sciences and Humanities Research Council (SSHRC) of the Canada-Insight Grant Program (No. 435-2020-0628) for supporting this research. Funding Information: The first author thanks the Finnish Work Environment Fund (No. 200224 ) for supporting this research. The second author sincerely thanks Prof. Charles D. Bailey of James Madison University for making his data available. Without it, this study would not have been possible. He also conveys his thanks to the Social Sciences and Humanities Research Council (SSHRC) of the Canada-Insight Grant Program (No. 435-2020-0628 ) for supporting this research. Publisher Copyright: © 2021Production and operations management (POM) uses learning curve (LC) models to determine the length of training sessions for new workers and predicting future task performance. Empirically validated LC parameters provide managers with quantitative information on the effects of the presumed factors behind the learning process. Previous studies considered LC to compose of cognitive and motor curves. Another widely acknowledged but only recently parameterized phenomenon in the POM field is interference, which assumes some loss of information or experience could occur over a learning session. This paper takes a logical step in this line of research by developing an interference-adjusted power LC model, a composite of cognitive and motor elements. This paper accounts for the decay of cognitive and motor memory traces from repetitions to measure the residual (interference-adjusted) experience and capture these phenomena. Three variants of the model are developed that assume power and exponential decay functions and an approximate version of the exponential one. Assembly data representing various forms of an individual learning profile have been used to test the fits of the developed models. In addition to those models, four potential models from the literature were selected for comparison purposes. The results show that the approximate model fits very well exponential learning profile. The findings highlight the confluence of the three phenomena in learning, component (cognitive/motor) learning, interference, and plateauing. (c) 2021 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )Peer reviewe