Abstract
The misalignment between curricula and industry expectations in Computer Science (CS) is a well-documented concern in Computing Education Research (CER) [1, 2, 3, 4, 11, 12]. Recently, this has been exacerbated by AI tools rapidly reshaping software development practices while academia slowly grapples with whether and how to incorporate them into curricula. Studying summer internship experiences offers a window into the growing academia-industry gap from the perspective of students. Although CER researchers have studied their experiences [6, 7, 8, 10, 13] and internships’ effects on their capstone projects [5] and study habits [9], little work examines internships as authentic learning environments in their own right. Furthermore, research capturing interns’ real-time perspectives remains limited, despite the fact that they are uniquely positioned to reflect on learning across both academic and industry contexts. To capture these perspectives, we conducted semi-structured interviews with six CS students both during and immediately after their internships. We inductively coded their responses to answer our research question: What do computing students learn during their summer internships, and how does it compare to their school-based learning? We found that internships overwhelmingly introduced skills, such as code review and navigating large codebases, that are structurally difficult to cultivate in school. As P6 explained, "I don’t know if the school can really adequately prepare you…[internships are] the only way you could learn how things go in the real world." Internships also introduced real-world stakes absent from school. For P5, this meant that her code was “actually” reviewed, since "actual clients are going to eventually use your code," and if a bug slipped through, P5 and her teammates would "take the fall." Many participants also mentioned the importance of being able to use AI while programming. For example, P2’s company expected employees to use an LLM to program, but she had little experience doing so, as AI usage was disallowed in her university courses. She determined that if she were “a little more accustomed to integrating with AI,” she’d be better able to keep up with the expected pace. This suggests that inconsistent curricular policies around AI tool usage may be widening the academia-industry gap in ways that CER has yet to examine. While all participants cited the value of their internships as sites of learning, we found that they experienced varied mentorship styles. Some were expected to contribute without much guidance, while others worked on teams that prioritized professional practices and observational learning over productivity. However, regardless of the team’s mentorship philosophy, participants spoke to what they believed school could not teach them. P4 noted that despite his team being unsupportive, "I learned something I wouldn’t learn in academia." Our preliminary findings argue for treating internships as a complement to, rather than a downstream outcome of, formal CS education. While incorporating industry-relevant skills into classrooms is important, aspects like working on a large team with large codebases cannot be easily replicated in educational settings. Ultimately, our findings emphasize CS departments’ responsibility to actively support students’ access to and success in internship experiences.
Funding Information
  • National Science Foundation (2313998)