General Project Guidelines · "Impact" Track · "Classroom" Track

This project track focuses on solving a real problem with AI and deploying your project to users. You will be required to build and deploy a working, end-to-end system (e.g. a tool, app, website) and run a thorough user study with members of your target audience. There are no restrictions on AI use for coding in this track (see AI Policy below).

The goal is to build something that users derive real value out of (not just potential value), i.e. people should find your project genuinely useful and it should have a positive impact on their lives. For example, it's not enough to build a restaurant recommender, have a few people try it out, and rate it 1-5. A tool that cuts the time someone spends ordering groceries from 20 minutes to 5 minutes demonstrates much clearer value.

AI should be at the core of the system, but you are not restricted to methods taught in the class. Your project can span anything from traditional ML methods to modern foundation models and AI agents.

It takes time, effort, and patience to build something that actually helps people, so we recommend starting early!

tl;dr: AI coding tools are allowed (and encouraged!) in this track. In both tracks, you will not be allowed to use AI to write your proposal or reports.

• Code: Use of AI coding tools (Cursor, Claude Code, Codex, Copilot, etc.) is fully allowed in this track. These tools represent the modern reality of software development. You can use AI to help write, debug, and understand code, and to explore different implementation approaches. Be careful: "vibe-coding" can often lead you into situations where you don't understand your implementation well enough to debug it.
• Writing: You may not use AI to write your proposal or any reports. This policy is consistent in both project tracks. However, you are welcome to use AI for brainstorming ideas, finding relevant literature, providing feedback on your writing, and formatting your reports. You take full responsibility for the contents submitted under your name, including content generated by AI tools that could be construed as plagiarism or fabrication (similar to ICLR guidelines).

AI-Use Disclosure Statement

Every writeup (proposal, progress report, and final report) must include an AI-Use Disclosure Statement, modeled on the ICLR policy on the use of large language models. In this section, you must describe, in roughly a paragraph or two, the precise role of any generative AI tools in that stage of the work, including ideation, writing, code, collecting or labeling data, running experiments, and analyzing results. As at ICLR, this section may appear in the appendix and does not count toward the page limit, and AI tools are not eligible for authorship. The statement will not be graded on its own, but omitting it or failing to disclose significant AI use may lead to substantial grade deductions.

1. Proposal (2 pages max). The proposal is due Oct 23. It should include the following items.
   • Problem Statement and Motivation: What problem will your project try to solve and who is it designed for (target audience)? What is the impact or value your project will have for your users? Make sure that the scope of the project is not too narrow or broad. For example, building an AI app for civic engagement is too broad, whereas a tool that summarizes city council agendas and flags items affecting your neighborhood is more reasonable.
   • Existing Solutions: What existing tools or systems attempt to address this problem or similar problems, and where do they fall short? Think about the limitations of existing solutions, e.g. missing features, poor reliability, or being not designed for your users. How will your system address these gaps?
   • User Research: Reach out to a handful of people in your target audience to get a sense of what they want and their thoughts on this problem. Is this a problem for them, and what potential solutions would they want? Pitch your project ideas and report their feedback.
   • Methodology: How will you approach the problem? Which AI methods or models would your system use, and why are they a good fit for your problem? What would be the main deliverables (e.g. an app, website, tool) and components of your project?
   • Evaluation Plan: What are the key goals of your project? How will you know if you've succeeded (outline clear success metrics, e.g. how much time this saves for your users)? We will ask you to run a user study as part of the final report. Start thinking about who your target audience is, how you plan to reach them, and how you plan to collect their feedback.
   • Description of the Challenges: What are the main challenges of building this system and getting people to use it? How might you address those challenges?
   • AI-Use Disclosure Statement: See instructions above.
   
   Note that you can still adjust your project topic after submitting the proposal, but your progress report (described next) should be on the same topic as your final report.


2. Progress report (4 pages max). The progress report is due on Nov 20. By this point you should have a working prototype or demo of your system, so you can get feedback from some initial users. Your report should include the following items.
   • Introduction: A specification of the problem the system solves and for whom, in one paragraph.
   • Main Approach: What does your system do? What are the key features, and how do they address the problem for your users? Walk us through a concrete example of how a user would use your system. Describe the AI methods or models you're using, as well as the key design decisions you've made. At this stage, we don't expect all your design decisions to be finalized yet, so this doesn't have to be extremely detailed.
   • Prototype: Screenshots or a short video of what your system can do so far, demonstrating the key workflow / ideas. It doesn't need to be deployed yet; a local version or a mock is fine.
   • Initial User Testing: Have a couple of people you're building for try out your prototype. Describe what happened when they tried it and what feedback they provided. What does the prototype already do well, where does it fail, what surprised you, and what will you change going forward?
   • Success Criteria: The concrete things you will measure and report in the final report to show that your system actually helps your users (e.g. how much time it saves them, or what made it valuable or useful to them), updated from your proposal based on your initial user testing.
   • Timeline: A timeline for completing your project. Provide a list of TODOs, e.g., wrapping up dev work, launch plan (e.g. launching it as an app / website / tool / etc.), user feedback study design and recruitment, final report writeup, and assign dates to each one.
   • Code: A link to your GitHub repository with your code so far.
   • AI-Use Disclosure Statement: See instructions above.
   • References: Please include a reference section with properly formatted citations. Not included in the 4 page limit.


3. Progress check-in with CA mentor (15-minute meeting). The check-in should be completed by Fri Nov 20. To provide more mentorship for the final project, we introduce a progress check-in between project groups and their assigned mentor. You will have a 15-minute meeting with your mentor to discuss the progress of your project, clear any confusion, and get advice. Your assigned CA mentor will reach out to give information on how to schedule the meeting for check-in.


4. Final Video Demo (5 minutes): The final video is due on Dec 4. The final video is a 5 minute demo of your system. In the video, you should briefly describe who you built this for and what problem you are trying to solve. Then, give a demo of your system being used, highlighting the main AI methods and aspects at the core of your system. You should focus on conveying the experience of the system, rather than an extensive tour through all the features. Finally, share some highlights from your user feedback study, as well as briefly discussing some challenges and key takeaways/lessons learned while doing your project.


5. Final Report (5-10 pages max): The final report is due on Dec 4. The structure is fairly similar to the progress report, but now we want to see your full system (not just the prototype), a comprehensive user feedback study, and a detailed analysis of the results and impact of your project.

   Your project should be a real, working system that people can use (not a prototype or a mock).

   Your report should include the following items.
   • Problem and Motivation: A summary of the key problem you solve, who you built it for, and what happened when people actually used it.
   • Existing Solutions / Related Work: What existing tools / systems address this problem, where do they fall short, and how does your system fill the gap?
   • Main Approach: What does your system do and how does it work? What are the key features, and how do they address the problem for your users? Describe the AI methods / models you used and the key design decisions you made along the way. Include screenshots of your system.
   • User Feedback Study: Have at least a few people you're building for use your system. Describe who you launched to and how, what they did on the system, and summarize what feedback they provided. Include quotes from users, organized into high-level themes, and discuss any unexpected or undesired behaviors. What worked well, what could be improved, and what did you change based on user feedback? See the FAQ below for tips on how to run a good user feedback study and what to report.
   • Results and Analysis: What impact did your project have? Report on the success criteria from your progress report, focusing on the real value your users got from your project (not the potential value). Back this up with concrete measurements, e.g. how much time your system saved your users compared to before.
   • Discussion and Future Work: Reflect on what worked, what didn't, and why. What challenges did you run into, and how would you improve the system if you had more time?
   • Ethical Considerations: An explicit discussion of any potential safety or ethical issues, such as the implications of the project, the use of the data, the actions the system is permitted to take, and potential applications of your work. For each issue, propose at least 1 possible mitigation strategy. You are not required to implement it.
   • Code: Please include a link to your GitHub, with a README containing instructions on how to run your code, plus a link to your deployed system (or instructions on how to install / use it). For private repos, make sure to communicate this with your mentor and get their GitHub ID to add to your repo.
   • Authorship statement: A brief statement explaining how the individual authors contributed to the project.
   • AI-Use Disclosure Statement: See instructions above.
   • References: Please include a reference section with properly formatted citations (any format of your choice).

   Note: you can have an appendix beyond the maximum number of allowed pages with any figures, screenshots, examples, or survey questions that you need. References do not count for the page limit.

We will give feedback on the following dimensions:

• Task definition: Is the problem statement precisely defined and is the motivation for the task clear? Does the project address a need for your users? We will reward projects that are extra thoughtful about how to use AI to help others in their community.
• Approach: Was the system design described clearly and well justified? Is AI the core of the system, and do the methods or models you chose fit the problem?
• User Feedback Study and Analysis: How thorough was your user research, feedback, and testing? Have you explained clearly who the system was launched to, collected feedback systematically, and reported concrete qualitative and quantitative results? Does your evaluation show the real value your system gave your users, not just that people tried it or liked it? You are not graded on the number of users, but on the thoroughness of your evaluation, as well as how well your results are documented and presented.
• Demo quality (final video): Does the system run and behave as described? Is it a real, working system that people can use (not a prototype or a mock)? Is it well polished, with a good user experience? Is the video a clear, compelling walkthrough of the system and core user experience?
• Ethical Considerations: Was at least one ethical concern or potential societal risk identified and explained within the context of the project? Was a strategy to mitigate each concern or risk identified (does not need to be implemented)? Both questions must be answered to receive full credit.

Since AI coding tools are unrestricted in this track, we will expect more from the projects compared to the "Classroom" project track, e.g. deploying to real users and collecting their feedback.

We recommend picking a well-scoped project that solves a real problem you care about, for people you can actually reach out and gather feedback from. Some ideas:

• Tool that helps manage logistics for a specific small business (e.g., tracking orders, invoices, and inventory)
• Website or app that keeps you updated on your local community or government and helps you find ways to get involved
• Meal-planning agent that helps you plan meals and order groceries
• Study tool that turns lecture recordings and slides into notes and practice questions
• Research assistant that discovers and summarizes papers based on your interests
• Campus-event aggregator and recommender

Frequently asked questions for this track

For questions that apply to both tracks (using the same project for another class, choosing partners, and writing the ethical considerations statement), see the general project page.

How do I run a good user feedback study, and what should I report?

Aim for at least a few participants who are not in your project group. We strongly encourage you to seek out the members of your community that you are specifically building for. Participation must be voluntary, and you should get your participants' informed consent before the study. You can modify this consent form from CS194H.

During the study, give the participant task directions that tell them what they are trying to achieve (not how to do it). Ask them to think aloud, and keep a log of their reactions, both positive and negative. Right after the study, ask them for feedback. You should carefully design a feedback form with the key, concrete questions you want them to answer, as well as an option for open-ended feedback.

In your report, describe the participants, the tasks, the procedure, what you measured, and the results. Include the questionnaires or surveys you used (e.g. in the appendix), and a writeup of what changes you made and why, based on what you learned from your study. Use quotes from your participants to support your findings.

[Sources: CS194H and CS147]

Will I be graded on the number of users?

We require you to gather feedback from at least a few people for this project and to run a user study with them. You will not be strictly graded on the number of users. Quality matters more than quantity: if 2 people use your system every day and get real value out of it, that's great!

If you launch and do not attract many users, this is still useful data. Work with your mentor to iterate on your design based on the feedback you do get, or to adjust your goals, and write about what happened in your report: what you tried, what you learned, and what you would try next.

[Sources: CS329Z and CS278]