AI Policy Chatbot

During the summer in 2024, I had the opportunity to complete an AI internship with SureStart, a company that trains educators and schools on how to use AI ethically and securely to improve their systems and define their own custom AI policies. I spent a few weeks researching and presenting on AI in education, examining case studies and research, AI tools used in schools, existing school policies with different stances, and best practices for schools when using AI or developing policy set by UNESCO and other organizations.

My main project was to develop an AI tool that could generate a rough draft of an AI policy based on that research, customized to the user's school information, personal stance on AI, approved tools and processes, and other input. I practiced prompt engineering in order to get the output needed for this purpose, and developed a chatbot in Python using the OpenAI API that runs in the terminal. After taking user input, the chatbot generates each section of the AI policy and updates the section based on user feedback, then asks for overall feedback on the whole policy and checks for consistency issues in tone, facts, or other content. Finally, it outputs the completed policy for the user.

Late in the project, I was asked to add a web interface, and after some research, found that the best solution was converting the program to JavaScript so it could run directly in the browser. This caused some issues, as the OpenAI Python API and JavaScript API were fundamentally different, resulting in a longer development process than expected. Although the project came with challenges, it taught me a lot about AI development and the challenges and rewards of working on real world projects. I'm sure the tool will be beneficial to SureStart's mission of helping schools take advantage of the power of AI.

To protect SureStart, I will not be sharing the code for the chatbot, but if you'd like to learn more about the project, please feel free to contact me.

Thanks for checking out my project!