Call for Proposals: Tools for Learning with AI

We are collectively building tools to help students and instructors at UNC leverage AI to enhance learning in UNC courses. The feature your team chooses to work on will be inherently experimental and avant-garde as we are building on the frontier of what is possible. Your work will be situated in a platform organized around courses, like Gradescope. You can imagine it as: learnwithai.unc.edu

In class on Monday, March 23rd, you all brainstormed ideas for tools to assist students in learning in UNC courses. Your table's ideas do not need to be final, but they are a reasonable starting point. You and your peers are the target user group for this work, in addition to instructors and TAs, which is critical to us being able to circumvent an important part of user-centered design which is understanding and interviewing the target user. A key difference between this course and COMP523: Software Engineering is in 523 you work with external clients and you and your peers, typically, are not the client. This will allow us to move faster in the remaining weeks of the semester, while still having a deep understanding of the user.

In the learnwithai.unc.edu platform, there are two distinct types of features you can attempt to take on:

1. AI Tools for Students or Instructors

These tools exist and are available in the context of a course for use at any time. Perhaps this is something like a Syllabus Chat Bot or a Practice Question Generator, but you should think creatively and broadly about what your group is motivated to build. This tool may be for students only, may be for instructors only, and perhaps it is useful to both. Furthermore, this type of tool may have some kind of configuration that the instructor is responsible for completing prior to it being usable by students, for example, a Syllabus Chat Bot would require configuring the syllabus content. However, unlike Learning Activities (which follows), students using the tool does not result in a submission. Using these tools should result in persisted artifacts that the user can refer back to (e.g. an Instructor Joke Generator produces jokes about a topic that are saved for later comedy).

2. Learning Activities for Students created by Instructors

Think of these like assignments that instructors or TAs design and create for students who are responsible for completing a recorded submission for. However, we are not worrying ourselves with grading concerns. Enhancing learning are our activities' primary goal. A simple example of an activity might be something along the lines of, "Code Describer" where an assignment might feature a prompt and a snippet of code the instructor puts together and then every student's assigned activity is to describe what the code does in English, in their own words, and AI provides constructive feedback. The submissions are recorded such that instructors are able to know who completed the activity. Many different instances of the same activity are likely to be assigned throughout the semester, just like reading GRQs in COMP423. Other ideas include things like, "quiz prep" or an open ended question worksheet, all of which could provide near-instant feedback from an AI. You can get much more creative with this than these ideas, but at the core an "Activity" is one where course staff creates and assigns it and students create submissions to it, which should get quick feedback of some kind, and course staff are able to see student submissions.

Scope and Relative Difficulty

Your team will focus on one tool or activity for the remainder of the course, which has only 5 weeks remaining. Generally, creating a Tool is less difficult than creating a Learning Activity. Your team should negotiate how confident you are feeling in the course thus far, as well as how much experience you collectively have on the frontend of web application development, to choose a feature and scope that feels realistic to your strengths. There is no grading upside one way or another, full credit can be earned working on a Tool or a Learning Activity. You should choose a feature that feels at the limit or just beyond your team's comfort zone because it will result in both learning more and a project that is more interesting to have in your portfolio of work to discuss with employers. Plus, with modern AI engineering tools, you will have opportunities to learn while creating.