Logo image
Back
Beyond e-learning in the age of artificial intelligence: a convergent mixed-methods investigation into optimizing workplace development using universal design for learning and generative pedagogical conversational agents
Dissertation   Open access

Beyond e-learning in the age of artificial intelligence: a convergent mixed-methods investigation into optimizing workplace development using universal design for learning and generative pedagogical conversational agents

Tyler Dean Creek
Doctor of Education (Ed.D.), Drexel University
Jun 2026
DOI:
https://doi.org/10.17918/00011476
pdf
Creek_Tyler_20262.51 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Artificial intelligence Bloom's 2-sigma problem Generative pedagogical conversational agents Sales engineers Universal design for learning Workplace learning
Organizational upskilling needs are rising, yet traditional e-Learning is time-consuming and poorly suited to the rapidly evolving needs of knowledge workers, such as sales engineers in information technology. This convergent mixed-methods study compared a course assisted by a generative pedagogical conversational agent (GenPCA) with a traditional e-Learning course, measuring assessment scores, learner-perceived universal design for learning (UDL) expression, and development cost efficiency. Grounded in UDL and mind, brain, and education science, this study situates the results within Benjamin Bloom's 2-sigma problem. Forty-five sales engineers completed the course, a knowledge posttest, and a postsurvey of Likert-scale and open-ended items, with development time tracked for each modality. Quantitative data were analyzed using t tests, Cohen's d effect sizes, and Mann-Whitney U tests, and the instrument's reliability was confirmed with Cronbach's alpha and Spearman-Brown. The qualitative data were analyzed through two-cycle coding, with the strands merged during interpretation. Assessment scores did not differ between the groups. Both cohorts arrived with prior expertise, and their scores clustered near the top of the instrument. This ceiling effect compressed the between-group variance, restricting the range needed to detect relationships among UDL perceptions, GenPCA didactical factors, and outcomes, so the null findings were traced to a single boundary condition. UDL expression favored the GenPCA modality, with large effect sizes across all three UDL networks: engagement, representation, and action-expression. Development effort favored GenPCA, requiring 8 hr against 119 hr for the traditional build, a roughly 15-fold reduction for this single course. Three contributions extend the work, including the inductive finding of aspirational convergence, a shared aspiration of learning affordances voiced across both modalities, the framing of the ceiling effect as a boundary condition on Bloom's 2-sigma, and the UDL-GenPCA Experience Inventory, a prototype instrument of UDL and GenPCA didactical factors with strong internal consistency. The GenPCA modality showed greater expression and required less development time. The assessment comparison remained inconclusive because the ceiling effect rendered any true difference statistically invisible. That boundary condition applies to tenured engineers with foundational content, not the earlier-career practitioners who stand to gain the most. Designing and measuring for those learners is the work this study leaves to the field. Keywords: generative pedagogical conversational agents, universal design for learning, workplace learning, sales engineers, Bloom's 2-sigma problem.

Metrics

1 File views/ downloads
2 Record Views

Details

Logo image