This study analyzes patterns of responses from a series of student surveys that predict student success in a CS1 course at a major research university in the United States. The research identifies statistically significant effects on final grades from self-reported course interaction perceptions and preferences, providing insights into designing future studies and potential early interventions.

Research opportunities offer students at the master's level a chance to apply their knowledge, create projects for their portfolio, and to gain an understanding of the research process in preparation for potential doctoral studies. The traditional structure of these opportunities, however, is not trivial to translate to online programs with asynchronous delivery. As online programs grow, it is important to examine ways that the traditional benefits of these opportunities can be extended to a broader range of students. In this poster, we discuss our experience building the infrastructure and running early efforts at reaching this goal. We discuss several dedicated courses and seminars developed to offer larger scale opportunities for students to pursue research. We also discuss efforts in active development to provide better infrastructure support to reduce the friction and complexity of pursuing research at scale.

The research focuses in academia are typically determined from the top down, with professors focusing on projects that align with their existing lab or can be readily supported by grants. This approach is pragmatic, but this focus may not align with the bottom-up interests of the available student body. As part of a broader effort to expand the available research opportunities in a graduate program in computer science, this work focuses on collecting data on what is motivating student interest in research and what specific fields students desire to study with the intention of using the results for decision-making about how to allocate resources to best match student interests. This work reports on the responses of 143 graduate students in computer science at a major research institution in the United States in a course-based program where access to research opportunities is not guaranteed. It examines both the fields of particular interest and the self-reported motivations leading these students to attempt to seek out research opportunities.

Past research on the ways that students evaluate their courses has focused largely on how those evaluations relate to the specific course instructor. This research examines a set of data from a public site where students unofficially rate the courses in a very large online graduate program operating at scale. We examine the relationship between the unofficial scores students give to their classes with data on enrollment trends over time and the assessment strategies used within the courses themselves to examine additional actors that shape the ratings students choose, as well as how they use those ratings to choose what courses to take in the future. We find several different notable relationships: reviews in this context are largely impervious to the extreme response bias prevalent on other review sites; review content does not appear to significantly influence enrollment trends; more difficult classes tend to receive more favorable ratings overall, although individual students do not rate difficult classes more favorably; and project-based classes are perceived by students to be less difficult.

Students have developed their own platforms for sharing their evaluations of courses over the Internet. These are typically focused on the needs and experience of students in traditional undergraduate programs, but the rise of online programs operating at scale has made it practical for students to develop such a platform dedicated to their particular program. We have used a survey to gather information from students in such a program at a major research institution in the United States. Through this data we explore how many students are using the site, how they use the information, and also how often and why they write reviews. The ultimate goal is to gather information that could help students to decide how to critically assess such reviews and successfully use them to make better decisions.

One of the major criticisms of efforts towards offering education at scale has been the Trap of Routine Assessment, the risk that student assessment will suffer from becoming excessively simplified in service of automation and scale. In this research, we examine the ways that students in an at-scale graduate program in computer science were assessed during their degrees. The program in question has scaled to over 10,000 students in only a few years, but awards a traditional Master's degree, providing the opportunity to investigate whether scale was achieved by transitioning to more routine assessment or by bringing scale to traditional strategies. To do this, we investigate the syllabi of 52 classes offered through the program to identify the types of assessments used, and we survey teaching teams for their approaches to evaluating these assessments. We merge this data with historical enrollment data to gain an overall summary of the kinds of assessments and evaluations received during their degrees. We ultimately find the program's scale has been managed by scaling up traditional assessment and evaluation strategies as the majority of grades are generated by human teaching teams based on projects and homeworks, with a relatively smaller portion generated exclusively by automated evaluation of exams.

Ensuring that students are fully informed about course content and policies is always a challenge, but online education at scale adds additional complications. In this paper we present observations about the place of the syllabus in education at scale, based on the actual syllabus documents from 48 courses in a Computer Science Master's degree program offered online and at scale. On the basis of these observations, we offer preliminary recommendations for factors that instructors should keep in mind when they compile a syllabus for similar courses.

Demand for CS education has risen, leading to numerous new programs, such as the rise of affordable online degrees. Research shows these programs meet an otherwise untapped audience of working professionals seeking graduate level CS education. In this study, we examine the motivations for enrollment among students in one such online MSCS program. Based responses to an open ended question, we develop a typology of motivations, including goals (e.g. career transition), opportunities (e.g. enrolling without taking time off work), and assurances that their goals will be met (e..g the program's accreditation). We then issue a closed survey question to a new group of students to further explore these motivations. In this paper, we discuss both aggregate and demographic trends in motivations, including the different motivations of men and women and what they imply about the program's impact on the gender divide in computing. We also examine older students' tendency towards intrinsic motivation to pursue an MSCS degree.

Effective collaboration between humans and artificially intelligent agents will require that the two are equipped to build a sense of mutual understanding with each other. When humans have an intuitive understanding of the motives and intentions of other humans, it is known as Theory of Mind. My work revolves around designing artificial intelligence to leverage this capacity to improve human collaborations with artificial agents.

Jill Watson is our name for a virtual teaching assistant for a Georgia Tech course on artificial intelligence: Jill answers routine, frequently asked questions on the class discussion forum. In this paper, we outline some of the ethical issues that arose in the development and deployment of the virtual teaching assistant. We posit that experiments such as Jill Watson are critical for deeply understanding AI ethics.