Graduation Semester and Year
Spring 2026
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Computer Engineering
Department
Computer Science and Engineering
First Advisor
Dr. Cesar A. Torres
Second Advisor
Dr. Ming Li
Third Advisor
Dr. Allison Sullivan
Fourth Advisor
Dr. Christine Dierk
Abstract
Tacit knowledge, knowledge that practitioners possess but cannot fully articulate, underlies procedural skills across virtually every domain of human practice. Online tutorials have become the primary medium through which people acquire procedural skills, yet the aspects of a task that matter most are often precisely those that demonstrators never name. Michael Polanyi's proximal-distal framework described how tacit knowledge resists explicit formulation, grounded in settings where a single demonstrator conveys knowledge to one learner (one-to-one) or to a group of learners (one-to-many). The contemporary tutorial ecosystem operates differently: demonstrators are a distributed population of contributors who produce knowledge artifacts like video tutorials and written guides, and learners navigate these artifacts through different interpretive lenses shaped by their experience and prior knowledge. This dissertation presents the many-to-many model of tacit communication to account for this configuration, introducing the task space, the collective knowledge of a task as contributed by its demonstrators, as a central construct. Three bodies of work operationalize this model. First, to computationally extract and organize tacit information from the task space, I developed methods for identifying and classifying the materials, tools, and techniques encoded across tutorial collections, making it possible to investigate overlaps between 25 communities of practice and illuminate how tacit knowledge flows across them. Second, to capture the interpretive lenses through which learners perceive tacit information, I developed the Tacit Description Typology through a crowdsourcing study in which 70 participants produced over 600 descriptions of tacit actions. The typology informed three design concepts for augmenting tutorials: Tacit Layer, which overlay descriptions onto a tutorial to communicate tacit qualities that were not verbalized; Tacit Space, which annotated a single aspect of a task across its full range of approaches so learners could distinguish between positive and negative actions; and Tacit Localizer, which paired learners with peers who perceived, thought, or talked like them to support more intuitive learning partnerships. Third, to assess what learners already knew about a task space, I developed Tacit Repertoires, an AI-assisted pipeline for constructing per-learner repertoires of prior knowledge. The pipeline captured self-reported knowledge availability, demonstrated understanding, and the learner's reconciliation of discrepancies between the two. These repertoires enabled repertoire-tuned tutorials that augmented tutorial content based on each learner's assessment performance. These contributions provide a computational framework for organizing the knowledge a task space encodes, formalizing how learners interpret tacit information, and assessing what each learner already knows, enabling personalized instructional support that bridges the gap between the knowledge tutorials make available and the knowledge each learner still needs to acquire.
Keywords
Human computer interaction, Tutorial, Tacit Knowledge, Crowdsourcing, Natural language processing, Generative artificial intelligence
Disciplines
Other Computer Engineering
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Endow, Shreyosi, "Beyond Proximal-Distal: A Many-to-Many Model for Tacit Knowledge Transfer in Tutorials" (2026). Computer Science and Engineering Dissertations. 4.
https://mavmatrix.uta.edu/cse_dissertations2/4