To develop a framework for designing recommendations to improve developer behavior, I analyzed two different techniques to explore what makes an effective recommendation and motivate the need for a new approach. Research shows that peer interactions, or the process of discovering tools from colleagues during normal work activities, is the most effective mode of tool discovery. However, studies also show that opportunities for peer interactions in industry are declining. Thus, I developed the naive telemarketer design as a baseline approach for designing automated recommendations from bots. These preliminary studies provide insight into characteristics of recommendations for increasing adoption of useful development tools and practices.

To improve the effectiveness of automated recommendations for developers, I introduce developer recommendation choice architectures- a framework that uses concepts from nudge theory to design automated recommendations for increasing the adoption of better behaviors and practices. This framework consists of three principles: 1) actionability, 2) feedback, and 3) locality. The preliminary evaluation shows software engineers significantly prefer actionable recommendations over static approaches.

To evaluate the developer recommendation choice architectures framework, I first examined the GitHub suggested changes feature which incorporates all of the framework principles. This feature allows reviewers to make specific recommendations on lines of code to improve pull requests during reviews, and lets developers automatically apply, reject, or edit changes. To analyze suggested changes through the lens of my framework, I conducted studies comparing this feature to other systems of recommendation styles and empirically analyzing its impact on development practices on GitHub.

Finally, to further evaluate the framework and support the thesis statement I developed and evaluated a new tool called class-bot. This system incorporates developer recommendation choice architectures through using automated GitHub issues to encourage students to adopt better behaviors while working on coding projects by following the software engineering process (Requirements, Design, Implementation, Testing, and Deployment). The evaluation shows class-bot improved student productivity by increasing programming activity and preventing procrastination in addition to improving the code quality of submissions.

Flower (pronounced flow-er) is an Eclipse plugin that leverages powerful program analysis techniques while maintaining a minimalistic interface to assist developers in effectively navigating program flow. Our tool enables developers to simultaneously navigate control flow and data flow within the Eclipse Integrated Development Environment (IDE). Some of the features include clicking on any variable or parameter to automatically invoke the tool and highlight everywhere the variable is used and displaying links in the code as well as boxes directly above and below the editor to method calls and declarations where the selected variable is modified or referenced.

Computer science education faces a unique problem where enrollment in introductory courses is drastically increasing while the field's retention rate is decreasing. Programming assignments and grades play a major role students losing interest in the major. This project studies using concepts from automated program repair to improve CS education by providing valuable feedback and automatically grading assignments for beginner programmers. The results suggest our patch grading algorithm is effective for improving grades on introductory coding assignments.

This project focused on using the Alice 3D block-based programming language to integrate computer programming concepts and computational thinking into K-12 education, specifically middle school math. For this work, we organized Adventures in Alice Workshops for K-12 teachers interested in incorporating computational thinking and programming concepts into their courses, hosted an Alice Activity Day for students, created and mapped tutorials to standards for middle school students, and presented this work at a variety of venues for students and research presentations. This research was completed for the Duke University Computer Science Undergraduate Research Fellowship (CSURF) program with Dr. Susan Rodger.

For this work, we examined ways to integrate robotics into society and improve the interaction between robots and humans. We explored using the Robot Operating System (ROS) to program iRobot Create robots and AR Drone Parrot Quadricpoters. Our final project consisted of designing and creating SwagBot, a delivery system to transport objects between users. This project was completed as part of the Computing Research Association Distributed Research Experience for Undergraduates (DREU) program Dr. Chad Jenkins and the Brown University Robotics group.