Empowering State-Level Multilingual Education Leaders: Harnessing Generative Artificial Intelligence for Program Administration

Author

Geqigula M. DlaminiFollow

Date of Award

2026

Document Type

Dissertation

Degree Name

Doctor of Education (Ed.D.)

Department

Leadership and Innovation

First Advisor

Robert Calvert, Ph.D.

First Committee Member

Allison Rowland, Ed.D.

Second Committee Member

Martin Martinez, Ed.D.

Abstract

This study aimed to understand how generative artificial intelligence could empower program staff at the California Department of Education (CDE) who support state-level multilingual programs by providing training and opportunities to review existing practices and identify promising use cases. With limited fiscal and personnel resources, state program staff had to rethink and intentionally consider how to enhance their own capacity to keep program momentum and support practitioners in the growth area of multilingual education.

With empowerment theory as the theoretical framework, this qualitative, participatory action research study involved 11 members of the Multilingual Support Division at the CDE. These co-researchers received state-approved training and joined a community of practice to explore the use of generative AI. Over the course of a month, they engaged in a sprint and conducted action research cycles of planning, action and observation, and reflection to determine whether their hypothesized use cases aided in their assignments. Data collection methods included three surveys during the study and a follow-up survey three months later. Additionally, recordings and transcripts from the community of practice meetings, as well as a researcher's journal, were used in data analysis, providing rich additional context for co-researchers' experiences. The method for data analysis was Braun and Clarke’s (2020) reflexive thematic analysis. Their data analysis framework emphasizes the researcher's perspective as a crucial element in the process. To address the research question, the analysis focused on themes of empowerment (both psychological and structural), the utility of generative AI in the public sector, and whether contextual integration (conceptually) could be effective for this purpose.

The findings showed almost all co-researchers felt more empowered in their work after participating in the study and with generative AI. Other findings included evidence of psychological (individual) and structural (organizational) empowerment, as well as barriers to empowerment and the implementation of generative AI. The data revealed how co-researchers’ attitudes, confidence, and use of generative AI evolved, and how their exploration yielded insights into use cases that supported state-level multilingual program administration, such as research, content generation, document evaluation, communication enhancement, and process improvement. Evidence of structural empowerment included data supporting participatory decision-making, knowledge sharing, and collective learning. Data on obstacles to empowerment and the adoption of generative AI were also identified, including technical and platform constraints, as well as individual preparedness factors.

Practice changed as a result of the inquiry, and co-researchers shared their experiences internally and externally, further maximizing the potential impact. The findings of the study also exemplify public entrepreneurship. Recommendations comprised expanding access to generative AI, encouraging others within the department to undertake the same process, and conducting additional research with other state-level multilingual education agencies. Further research in this area may include studies within other state-level agencies and more advanced generative AI use cases.

Pages

158

Recommended Citation

Dlamini, Geqigula M.. (2026). Empowering State-Level Multilingual Education Leaders: Harnessing Generative Artificial Intelligence for Program Administration. University of the Pacific, Dissertation. https://scholarlycommons.pacific.edu/uop_etds/4322

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.