Research
Every intervention was instrumented. Baseline offerings were measured before any change, so the comparison is against the course as it actually was — not against a memory of it.
Publications
Work arising directly from the exemplar project and its immediate predecessors, as cited in the volume’s chapter bibliographies. Most recent first.
The last two are the curriculum guideline this project builds on — references [38] and [39] in the volume’s bibliography.
Presentations
The Edu* workshop series has been the central mechanism for building and sustaining a community of PDC educators, and is where this project reports its results.
The NSF/TCPP Workshop on Parallel and Distributed Computing Education. Began as the venue for early adopters and interested instructors to report experiences, receive feedback and discuss curricular and pedagogical issues. The project’s Peachy parallel assignments work appeared at EduPar-25 in Milan.
The complementary venue connected to the high-performance computing community, and home of the Peachy Assignments programme. The project’s visual unplugged activity paper was presented here in 2025.
Held with HiPC in India, extending the Edu* model internationally, and a venue for associated instructor training activities.
Technical symposium papers, tutorials, special sessions and the CDER booth, reaching mainstream CS educators. The codeless modules work appeared at SIGCSE TS 2026.
Evaluation methods
The instruments are set out in Chapter 0 § 0.3.6. Two were administered to students and four to the faculty themselves.
Each institution ran a baseline offering with no intervention before running the PDC one, so the comparison is within an institution rather than across them — different calendars, class sizes and student populations never have to be treated as equivalent.
A common instrument on experience with and attitudes toward PDC, given at the start and end of the course by both development and testing team instructors. Because every site ran the same items, results can be pooled.
After each lab and programming assignment: how much time it took, which days it was worked on, what was learned, the hardest part, what remained unclear, and how much it was enjoyed. This is what turns a chapter’s timing tables into something an adopter can plan against.
Applicants’ PDC background, current use of PDC in class, institution, courses and student population — alongside a subjective read on their attitude to putting PDC in lower-level classes.
Whether the exemplars were appropriate and effective, how likely PDC is to reach their other courses, and their first ideas about approach and intervention points.
Annual surveys during the summer working sessions counted students actually taught and how far PDC had been incorporated; a post-adoption survey asked whether the interventions will be retained, reduced or expanded, and what new directions followed.
Each testing team applied for IRB approval and set its own research questions, with data collection coordinated across teams from the start so the results could still be compared.
Matched pre/post responses with non-parametric Wilcoxon signed-rank tests and Cliff’s delta for effect size, reported per item; open responses coded thematically, so the numbers come with student reasoning attached.
Developing examples, finding curriculum insertion points, revising assignments and testing software took more than 100 hours across the initial development year — reported rather than glossed over, because adopters need to plan for it.
What the data shows
Full tables, figures and anonymized datasets are in each chapter’s results sections and appendices.
Large-section C++ infusion
Baseline Spring 2024 against four combined intervention semesters (Fall 2024, Spring 2025, Fall 2025, Spring 2026), with per-assignment agreement data and time-on-task for the Earthquake Tracker and OpenMP labs.
Java liberal-arts adaptation
Pre- and post-surveys were given every term the course ran, but the results are held back from this early release and will be added in a later update. What the chapter does carry is the qualitative account: which material was displaced, why it mattered less than expected, and the departmental conditions that made the change straightforward.
Survey results pending · Chapter 3 · p. 107
Structured Java adoption
Fall 2024 benchmark (N = 24) and Fall 2025 PDC-intervention (N = 33) cohorts compared item by item, with pre-to-post gains on four PDC measures including parallel processing, distributed computing and overall PDC understanding. Includes diverging Likert distributions of engagement items and thematic coding of open responses.
Codeless and low-code modules
Codeless modules evaluated across Fall 2024 and Fall 2025 with Wilcoxon signed-rank testing and Cliff’s delta effect sizes reported per item, with significance interpreted at p < 0.05.
n = 125 · Chapter 5 · p. 166
Visualization-first C++
Pre/post comparison of self-reported PDC experience and core C++ topic experience on a seven-point scale, plus attitudes and interest, and post-survey ratings of learning activities and outcomes.
Community-college adoption
Perception of learning gains in non-PDC topics stayed broadly flat while perception of gains in PDC topics rose in the intervention semester — the clearest available check that PDC content did not displace core learning.
8-week summer and 16-week semester runs · Chapter 7 · p. 202
Low-preparation activities
Pre-test and post-test performance by question, plus adapted ASPECT engagement results grouped by category for both the Flag Maker and Penny activities, across Fall 2024, Fall 2025 and Spring 2026.
Minimal-infusion model
A no-intervention semester compared against the Flag Maker intervention semester shows gains from pre- to post-survey across all basic programming categories in both, with comparable post-survey means — evidence that adding a short unplugged activity did not reduce students’ perceived progress in core CS1 content.
Shared activity families
The shared lesson is that PDC works best in CS1 when it reframes and enriches existing course goals rather than appearing as disconnected additional content — starting with intuition, using unplugged or visual activities before code, and keeping programming tasks lightweight.