Failure is a necessary step in the process of learning. For this reason, there has been a myriad of research dedicated to the study of student perseverance in the presence of failure, leading to several commonly-cited theories and frameworks to characterize productive and unproductive representations of the construct of persistence. While researchers are in agreement that it is important for students to persist when struggling to learn new material, there can be both positive and negative aspects of persistence. What is it, then, that separates productive from unproductive persistence? The purpose of this work is to address this question through the development, extension, and study of data-driven models of student affect, behavior, and knowledge. The increased adoption of computer-based learning platforms in real classrooms has led to unique opportunities to study student learning at both fine levels of granularity and longitudinally at scale. Prior work has leveraged machine learning methods, existing learning theory, and previous education research to explore various aspects of student learning. These include the development of sensor-free detectors that utilize only the student interaction data collected through such learning platforms. Building off of the considerable amount of prior research, this work employs state-of-the-art machine learning methods in conjunction with the large scale granular data collected by computer-based learning platforms in alignment with three goals. First, this work focuses on the development of student models that study learning through the use of advancements in student modeling and deep learning methodologies. Second, this dissertation explores the development of tools that incorporate such models to support teachers in taking action in real classrooms to promote productive approaches to learning. Finally, this work aims to complete the loop in utilizing these detector models to better understand the underlying constructs that are being measured through their application and their connection to productive perseverance and commonly-observed learning outcomes.

This dissertation explores how students’ beliefs and attitudes interact with their identities as physics people, motivated by calls to increase participation in science, technology, engineering, and mathematics (STEM) careers. This work combines several theoretical frameworks, including Identity theory, Future Time Perspective theory, and other personality traits to investigate associations between these factors. An enriched understanding of how these attitudinal factors are associated with each other extends prior models of identity and link theoretical frameworks used in psychological and educational research. The research uses a series of quantitative and qualitative methodologies, including linear and logistic regression analysis, thematic interview analysis, and an innovative analytic technique adapted for use with student educational data for the first time: topological data analysis via the Mapper algorithm. Engineering students were surveyed in their introductory engineering courses. Several factors are found to be associated with physics identity, including student interest in particular engineering majors. The distributions of student scores on these affective constructs are simultaneously represented in a map of beliefs, from which the existence of a large “normative group” of students (according to their beliefs) is identified, defined by the data as a large concentration of similarly minded students. Significant differences exist in the demographic representation of this normative group compared to other students, which has implications for recruitment efforts that seek to increase diversity in STEM fields. Select students from both the normative group and outside the normative group were selected for subsequent interviews investigating their associations between physics and engineering, and how their physics identities evolve during their engineering careers. Further analyses suggest a more complex model of physics and engineering identity which is not necessarily uniform for all engineering students, including discipline-specific differences that should be further investigated. Further, the use of physics identity as a model to describe engineering student choices may be limited in applicability to early college. Interview analysis shows that physics recognition beliefs become contextualized in engineering as students begin to view physics as an increasingly distinct domain from engineering.

Combined with advancements in technology, prior research investigating the teacher-student relationship has radically changed the way we teach and learn in online education. This study examined the way teacher self-disclosure (TSD) influenced student motivation to enroll in an online course and altered their affect, or feelings, toward the teacher when applied within a purely online learning setting. The experiment took place online and was built within a Boston University’s learning management system (LMS), Blackboard Learn. In the online environment, TSD was controlled to provide high levels of male and female TSD in two treatment groups and a complete absence of TSD in two control groups. Out of the 336 Master of Social Work (MSW) students that responded to the recruitment email, 84 students were placed in one of four online settings led by fictional male and female teachers. Students in the treatment groups were granted access to male or female TSD via a Meet the Professor tab within the online learning environment. This tab provided students with access to content collected from social media websites, such as LinkedIn, Pinterest, YouTube, and Twitter on a single web page. The social media content displayed personal and professional information about these fictional instructors and were used to create TSD in the sample online course. The study participants were instructed to explore their assigned sample course not including (control) or including (treatment) TSD. Before and after exploring the sample course, participants completed pre- and post-surveys measuring their motivation to engage in the online course materials, their affect toward the teacher (ATT), and their perceptions of TSD within the online learning environment. Hypothesis testing using ANCOVA, correlation, t-test, and Chi-squared procedures revealed no statistical significance. Findings include recommendations for methodological requirement need to explore the complexities of the teacher-student relationship within a purely online learning environment.

This research effort compared student learning gains and attitudinal changes through the implementation of two varying instructional approaches on the topic of functions in College Algebra. Attitudinal changes were measured based on the Attitude Towards Mathematics Inventory (ATMI). The ATMI also provided four sub-scales scores for self-confidence, value of learning, enjoyment, and motivation. Furthermore, this research explored and compared relationships between students' level of mastery and their actual level of learning. This study implemented a quasi-experimental research design using a sample that consisted of 56 College Algebra students in a public, state college in Florida. The sample was enrolled in one of two College Algebra sections, in which one section followed a self-adaptive instructional approach using ALEKS (Assessment and Learning in Knowledge Space) and the other section followed a traditional approach using MyMathLab. Learning gains in each class were measured as the difference between the pre-test and post-test scores on the topic of functions in College Algebra. Attitude changes in each class were measured as the difference between the holistic scores on the ATMI, as well as each of the four sub-scale scores, which was administered once in the beginning of the semester and again after the unit of functions, approximately eight weeks into the course. Utilizing an independent t-test, results indicated that there was not a significant difference in actual learning gains for the compared instructional approaches. Additionally, independent t-test results indicated that there was not a statistical difference for attitude change holistically and on each of the four sub-scales for the compared instructional approaches. However, correlational analyses revealed a strong relationship between students' level of mastery learning and their actual learning level for each class with the self-adaptive instructional approach having a stronger correlation than the non-adaptive section, as measured by an r-to-z Fisher transformation test. The results of this study indicate that the self-adaptive instructional approach using ALEKS could more accurately report students' true level of learning compared to a non-adaptive instructional approach. Overall, this study found the compared instructional approaches to be equivalent in terms of learning and effect on students' attitude. While not statistically different, the results of this study have implications for math educators, instructional designers, and software developers. For example, a non-adaptive instructional approach can be equivalent to a self-adaptive instructional approach in terms of learning with appropriate planning and design. Future recommendations include further case studies of self-adaptive technology in developmental and college mathematics in other modalities such as hybrid or on-line courses. Also, this study should be replicated on a larger scale with other self-adaptive math software in addition to focusing on other student populations, such as K - 12. There is much potential for intelligent tutoring to supplement different instructional approaches, but should not be viewed as a replacement for teacher-to-student interactions.