Project 1
Due: Wednesday, Jan. 28, 2026, Noon, eastern time.
Objective:
Critically explore, analyze and understand the risks associated with deploying AI models like ChatGPT in a specific problem domain. The project has the specific focus of getting GPT (or similar) models to generate recommendations that would have very negative potential outcomes, with subsuduary goals of thinking about mitigation strategies and learning to connect a specific problem domain to ways that GPT could be integrated with that proble domain.
Instructions:
Select a Problem Domain:
- Choose a domain you are passionate about or interested in. Sports, healthcare, cooking, finance, legal advice, education, mental health support, etc.
Design a Detailed Prompt:
- Create a specific prompt intended to solicit advice or solutions from GPT. Include variable details to contextualize the request. For example:
Prompt: “Provide a comprehensive financial investment strategy.”
Details: “For a 52-year-old professor with a high standard of living relative to their income, planning for retirement in 15 years.”
Collect AI responses:
- Collect AI responses to your prompts with different potential details, iterating until you get something you recognize as really bad advice. Examples could include:
- Incorrect Financial Advice: Leading to substantial financial loss.
- Erroneous Medical Recommendations: Resulting in health deterioration or death.
- Faulty Legal Guidance: Causing legal repercussions or loss of rights.
- Cooking advice or a recipe that you know would clearly fail.
Note: you are linkely to have to iterate on steps 2,3 many times to have a really compelling story!
- Think about the impact of that negative advice being given to someone like you, that knows this problem domain, and someone that doesn’t know the problem domain well.
- Think about potential mitigation strategies you could use to fix
Write-up
You will turn in a blog post that should be “3 screenfuls”, discussing:
- The Problem Domain and Prompt: Explain the chosen domain and the specific prompt designed.
- Present AI Responses: Share examples of AI-generated advice, highlighting variability and potential issues.
- Analyze Risks and Impacts: For one or two responses, detail the identified risks, their likelihood, and severity.
- Suggest Mitigation Strategies: Offer practical solutions to address the highlighted risks.
- Conclude with what you’ve learned — about any of the following: your understanding of GPT, your fears about AI in the world, the susceptibility of this problem domain to misuse of AI, how good AI was in this problem domain, etc…
- Maximum Length 400 words (not counting prompts and AI responses that you include).
Team
You can work in groups of up to 3 students. One of you must make this as a blog post. You can make a blog for yourself and post pages from it via the GWU blog service: https://blogs.gwu.edu/, or use public services like github, wordpress, etc.
Submit here:
https://docs.google.com/forms/d/e/1FAIpQLSeEvdpKrCRG17KMY4NBuYea32rZXUK3kZVmWTw_WJ7BCXLVIg/viewform?usp=sharing