Banca de DEFESA: Allan Ferreira

Learning retention in corporate training: evaluation of predictions in a longitudinal section

learning retention; self-efficacy; instructional design; support for training transfer.

The focus of this study is Learning Retention (LR), defined as the extent to which the changes that occur in an individual, measured in terms of knowledge, skills or attitudes, and which are the result of training, remain over time. The general objective of the study is to measure and evaluate prediction relationships for the LR variable, in a longitudinal perspective. To achieve the objective, we carried out four studies. In Study 1, we identified the current state of knowledge production on LR, through a narrative literature review, which encompassed 11 articles on LR, from peer-reviewed journals, published between 2017 and 2023 and available in 6 databases. We conducted the other studies by collecting data from workers at a financial institution, graduates of self-instructional, live online and in-person training. In Study 2, we constructed and investigated evidence of validity of an LR assessment instrument in corporate training. In Study 3, we constructed and investigated evidence of validity of an instrument for evaluating training characteristics associated with learning retention (TCLR). In these two studies, 10,960 graduates of the aforementioned training courses were subjected to the instruments developed and semantically validated for the assessment of LR (with 6 items) and for the assessment of TCLR (with 8 items), obtaining responses from 1,185 individuals (10.8% return); we carried out an Exploratory Factor Analysis with the Factor; and we calculated adjustment and reliability measures for the models (CFI, RMSEA, χ²/df, Cronbach's Alpha, McDonald's Omega). In Study 4, we verified the relationship between the predictor variables self-efficacy (SE), TCLR and support for training transfer (STT) and the criterion variable RA, in a longitudinal perspective. In Wave T1 after training, we measured RA and TCLR (N = 1,185). In Wave T2 (2 months after T1), the sample obtained in Wave T1 was submitted to instruments for assessing RA, SE and STT, reaching 520 respondents (43.9% return). In Wave T3 (2 months after T2), the sample obtained in Wave T2 was submitted to the instruments for assessing LR and STT, reaching 281 respondents (54.0% return). It is noteworthy that there is evidence of validity and reliability for all instruments used in Study 4. In each wave, we carried out logistic regressions (LogR) with Jamovi, defining the LR measured in the wave as the criterion variable, and as predictor variables all others measured (TCLR, SE, STT, LR measured in previous waves and sociodemographic and professional data). In each LogR analysis we also calculated adjustment and prediction measures (Deviation, AIC, BIC, R²M, R²CS, R²N, Accuracy, Specificity, Sensitivity and AUC). As a result of Study 1, a variety of LR assessment methods (what, how and when to measure) were identified. Variables associated with LR were also identified, such as SE and STT, as well as gaps, such as the lack of a comprehensive instrument for assessing LR and an instrument for assessing TCLR. The results of Studies 2 and 3 indicated evidence of validity for the single-factor LR assessment instrument (CFI = 0.998; RMSEA = 0.085; χ²/df = 2.86; Cronbach's Alpha = 0.970; McDonald's Omega = 0.970) and for the single-factor TCLR assessment instrument (CFI = 0.993; RMSEA = 0.072; χ²/df = 2.89; Cronbach's Alpha = 0.927; McDonald's Omega = 0.927). The results of Study 4 pointed to SE and STT as significant predictors of LR at T2 and T3; TCLR as a significant predictor of LR at T1 and T2; LR at T1 as a significant predictor of LR at T2 and T3; and LR at T2 as a significant predictor of LR at T3. In summary,  the results indicate that people with greater SE, with a perception of better STT, with greater LR (at T1 and T2) and with a perception of the existence of TCLR belong to the group with greater LR at T3. This work brought methodological contributions to studies on LR, by presenting an instrument for its evaluation and an instrument for evaluating TCLR, filling gaps identified in Study 1. There are also theoretical contributions, with the identification of the state of knowledge production on LR and with a better understanding of the relationship of this variable with the SE, TCLR and STT variables. As a research agenda, proposed based on the limitations found within the scope of this work and the gaps identified in Study 1, we suggest that future studies on LR be carried out in other fields of research, over longer periods, with experimental or quasi- field experiments and/or situational tests to measure LR.