Exploring AI Ethics in Computer Science & Engineering
Phil Barry
Computer Science and Engineering
College of Science & Engineering
Phil Barry is a teaching professor in the computer science and engineering department. He teaches courses in theoretical computer science and ethics and computing.
Please describe your assignment in a paragraph or two.
The assignment is part of an ethics and computing course. One theme in the course is ethics and generative AI. This scaffolded assignment has students use generative AI to produce and analyze a short persuasive paper on some topic in ethics and computing.
How does the assignment work in your class? What learning does it support?
The assignment has multiple purposes: (1) Having students use generative AI in a nontrivial way; (2) Having students work with ethical systems (such as utilitarianism and deontology) that are covered in the course; (3) Having students analyze the effectiveness of the ethical arguments made by the AI.
How have you developed and refined this assignment over time?
I've changed it slightly each time I've used it. Most changes have been toward greater specificity in expectations. And I'm still experimenting with the assignment. For instance, last semester I gave students (a choice of three) AI-generated papers, and had them focus on the analysis steps rather than have them come up with a topic and then use an AI to generate the paper. This was partially successful, and I will continue to experiment with other variations on the assignment.
What advice would you give to other instructors who would like to develop a similar assignment?
Be very specific in instructions to students. For example, the first time I used this assignment, many students did the minimal amount of work in coming up with a topic and having an AI produce a paper. The resulting AI-generated paper was often poor, which harmed the following analysis step in the assignment.
What do students say about the assignment? How have they found it valuable?
Feedback has been mixed. Some students did not engage with the assignment and did the minimal amount. Others found it very useful. For instance, one student said it was transformative for her, and convinced her of the usefulness of the class, not just the assignment itself. Other positive comments were about seeing explicit ethical reasoning, and about the chance to do an in-depth analysis of an ethics and computing topic.
What makes a student response to this assignment effective?
The best student work (1) used AI in a nontrivial way (e.g., used an in-depth, iterative prompting process to generate the paper), (2) did a deep rather than superficial analysis, and (3) used the AI as an "assistant" rather than over- or under-relying on the AI.
What We Like about Phil’s Assignment
Reflection is Unpacked: Phil's assignment culminates in a reflective paper, which can be challenging for students. A reflective paper requires students to engage in meta-cognition—thinking about one's own thinking. It also requires students to use their own work as the object of study.
When students struggle with reflective assignments, they may rely primarily on providing narrative and methodological details in lieu of analysis. In his assignment, Phil provides an “overview and purpose” section that emphasizes self-analysis by providing students with an explicit title—"What I Learned..." Phil reinforces this analysis with five specific learning goals, four of which stress "learning about" and one emphasizing reflection.
Read more about Designing Reflective Writing Assignments.
Assignment is Sequenced: Phil's assignment is very intentional in its sequencing of writing stages. One effective strategy for sequencing, which Phil uses here, is to scaffold and sequence assignments so they begin with concrete tasks before moving into reflective and abstract ones.
Read more about Scaffolding and Sequencing Writing Assignments.
Purpose and Requirements are Clear: Phil makes explicit the purpose for which students are writing and the goals they should achieve for each step of the assignment. Making expectations transparent and explicit helps students better understand the assignment and thus complete it more successfully.
In Step 1, by providing a checklist, Phil further elucidates characteristics of a successful recommendation, empowering students to proactively and independently refine their recommendations rather than dumping all their ideas and waiting for the instructor to tell them how to refine.
Clear AI Guidelines Support Student Thinking: While Step 2 in Phil's assignment requires students to use GenAI, Phil offers guidance throughout the assignment steps on how GenAI may or may not be useful in various ways. Phil stresses the role of the student's agency in making this decision. One useful consideration when providing guidance on AI use and writing is to ask students to do some thinking first before using a tool. Doing so helps to work against the "anchoring bias"—the tendency to rely too heavily on the first pieces of information we encounter—that AI tools can create.
Read more about Establishing AI Guidelines.
Writing Process is Emphasized: Along with providing carefully sequenced and scaffolded activities, Phil advises students on how to carefully document their work in order to support subsequent writing activities.
Read more about Communicating Multiple Roles for Writing in Your Course.
Length Requirements are Aligned with Writing Expectations: Many students will approach word counts as terminal points for stopping their writing. To counteract this, Phil stresses the need for efficiency, selectivity, and concision. Students might need to write more than the word count and then revise to meet the expectation.
Read more about ways to Support Students with Word Counts and Page Requirements.