UC Davis

It is becoming ever more challenging to address the changing landscape faced by California Community Colleges. Change agents strive to implement needed interventions but are often disappointed in the outcomes. However, the literature shows that the odds of successful change increase when interventions align with the organization’s culture. The problem is how do change agents identify elements of a college's culture?This study explores that question by using different methodological approaches to extract elements of college identity from documents that plausibly express key college priorities and values. Using the organizational identity framework, elements of college identity are conceptualized as latent themes embedded within artifacts produced by individual colleges - in this case, Institutional Self Evaluation Reports (ISERs) used for accreditation. To generate the broadest possible picture, my dissertation uses two different methods: a machine-learning content analysis method and a manual semantic analysis method. Using Latent Dirichlet Allocation, the computer extracted 5 latent topics from 25 Community College ISERs corpus. Concurrently using Dedoose, manual semantic analysis was conducted with vivo coding of the same corpus. Latent topics emerged by analyzing both sets of semantic topics separately. Several results emerged from the study: it confirmed that ISERs are a suitable data set for topic extraction, and both methods can extract latent themes. However, the themes did not overlap as much as anticipated. The automated method extracted topics with an institutional lens focused on specific processes, while the manual method used a broader perspective, focusing on outcome values and concepts. The results suggest that both methods, used together, can provide a more comprehensive picture than either method used alone. The study concludes with several recommendations that practitioners may find useful.