More than 3 years since the introduction of ChatGPT and other LLMs into the public sphere, AI has touched nearly every academic discipline. Students are using AI for things from writing help to tutoring, and faculty responses have ranged from absolute rejection to complete adoption. In conjunction with other efforts across campus, our goal is to offer one department’s process for determining which changes are needed for their curriculum.
One area of study that AI has dramatically affected is Computer Science (CS). CS graduates learn software development skills, and AI tools have proven very useful to enable developers to write code more efficiently. The challenge is to help CS students learn the fundamental concepts of CS without relying on AI, then to use AI to in their productivity to prepare them for a workforce that expects them to use AI tools. The Department of Computer Science at Boise State University formed a committee of faculty to strike a balance between those two goals. The following highlights some important take-aways from that committee’s discussions and recommendations that might be useful for other departments.
Specific Recommendations
The committee recommends that the department should create a student-facing summary document that explains the rationale for the upcoming GenAI integration that could be posted on the department website. We also recommend developing a short blurb that can be included in syllabi with a link to the summary document. Here are some samples: Student Summary and Syllabus Blurbs; also see the Boise State Center for Teaching and Learning’s website for samples and additional AI education support.
The committee emphasizes that the department should strongly encourage faculty members to experiment and gain experience with various AI tools. The university offers AI workshops that would fit the category of professional development. The department can consider holding 10-15 minute mini-workshops in department meetings where one tool or technique is demonstrated.
Determining Changes in Courses
In Computer Science, fundamental courses prioritize fundamental skill development (coding, analysis, problem solving, breaking down problems, debugging and testing, conceptual understanding) which are core skills that remain critical for students to be able to individually demonstrate. In order to better prepare students for the workforce, new AI skills (prompt engineering, critical evaluation and verification of AI-generated code, problem decomposition and top-down design, ethics and professionalism), the Department of CS has met to determine the needed changes.
Departments should consider what changes need to be made to their courses without sacrificing program or course outcomes.
Implementation Strategy
To help determine how AI could be integrated into each course (if at all), the committee broke down potential scenarios into 5 different integration “scenarios.” Each scenario is illustrated in the infographic and described below.
- Scenario 1: No GenAI
Students develop core competencies without AI assistance. GenAI use constitutes academic dishonesty.
- Scenario 2: Restricted GenAI
GenAI is permitted only for specific, instructor-defined tasks (e.g., explaining error messages, suggesting debugging approaches). Prohibited for code generation. Students must document AI interactions.
- Scenario 3: Guided GenAI
GenAI is treated as a development tool requiring critical evaluation. Students must verify, understand, and modify all AI-generated code. Emphasis on prompt engineering and output validation as learning objectives.
- Scenario 4: Open GenAI with Accountability
Students freely use GenAI but must demonstrate understanding through explanation, modification, or extension of solutions. Assessments focus on comprehension and application rather than initial code creation.
- Scenario 5: Full Gen AI Integration
GenAI usage mirrors professional practice. Focus shifts to system design, architecture, testing, and code review. Students evaluated on problem decomposition, tool selection, and quality assurance rather than manual coding.
Discussing Specific Recommendations
The committee then discussed a selection of courses (particularly lower-division, required courses) to determine which scenario fit best for each course, ensuring to align course objectives with the scenarios. The table below is an example of the courses, names, objectives, and recommended scenarios that were agreed upon by the committee.
The committee then provided a paragraph for each course to explain the reasoning. For lower division courses where AI use is discouraged, the committee recommended changes to assignments and assessment practices, where necessary:
…[w]e recognize it is a losing battle to prevent AI use. Therefore, assessments must be re-weighted and redesigned to make AI code generation without understanding a detriment to success, rather than a benefit. Producing working code, alone, will not be adequate to receive a passing grade in the course. Instead, at least half of all course points will come from demonstration of understanding through paper or proctored quizzes and exams without access to AI or other external resources. Programming assignments, then, are framed as learning and practice exercises to prepare for quizzes and exams. The focus of in-class and formative homework-level programming assignments may shift away from producing functioning code (although ground-up program development will remain for larger project assignments) and toward code comprehension, exploration, analysis, debugging, and comparison of alternative solutions. This approach demands understanding of applied coding concepts rather than rewarding producing code solutions as quickly and effortlessly as possible, muffling the siren call of LLM code generation.
The committee also made individual recommendations for the list of courses. One example:
CS 321 Data Structures – GenAI should be carefully managed to allow students an introduction to appropriate use while continuing to restrict it for foundational computer science concepts. For projects, GenAI would be used in S2 mode for the first individual projects and then students would transition to S3 mode for the large team project.
Helpful Examples
The committee then discussed examples of assignments, lessons, prompts, course content, etc., that could be used in courses. For the CS 321 example:
In-person or Online CS 321
See below for planned changes to the course for fall semester (most of the changes will be piloted this Spring):
- Assessments: Change all exams to be in-class. Add some form of interactive assessment for projects and decrease the weight for projects. In-person or proctored exams for the online section do have some challenges that would need to be resolved.
- Use a custom NotebookLM with all class notes, code examples, and project writeups to use as a study guide. See example here: CS 321 Notebook LM.
- Add AI exploration topics in each chapter.
- Ask students to hear the overview podcast produced by Notebook LM and find a flaw! (or hallucinations). Assign after the first three weeks.
- Students use CoPilot in ask/edit mode for p1, p2, and p3. Students will include a log of interactions and a reflection for each project. Create AI warmup project to be used before project 4 (in S3 mode)
- Instructions for AI in S2 mode: (add these to Notebook LM or any AI tool)
You are a tutoring system and not allowed to provide full code for any of the included projects. Instead, you should guide the user with hints, and explanations to help them complete their projects on their own. You may provide edits to their code but do not provide code for entire methods or classes or declarations of critical data structures.
Additional Resources
AI Use Assignment Intervention Rubric
Veritasium: What Everyone Gets Wrong About AI and Learning – Derek Muller Explains
https://www.chronicle.com/special-projects/the-different-voices-of-student-success/ai-to-the-rescue
https://www.chronicle.com/article/how-are-students-really-using-ai?