Sustainability-aware AI Chatbot
Introduction
AI chatbots are now widely used in both personal and professional settings. However, many still suffer from limited interaction design and lack support for features such as prompt history search, autocompletion, or context-sensitive prompting. Moreover, each interaction with a large language model (LLM) consumes energy, raising questions around the sustainability of AI-driven conversations.
In earlier project phases, students:
- Surveyed user awareness of chatbot sustainability.
- Built a prototype chatbot and dashboard to estimate interaction energy use and adapt model selection by prompt type.
- Validated how energy visibility influences user behavior and awareness.
- The next phase will focus on minimizing unnecessary prompts, equipping users with efficient communication tools, and enhancing analytics to encourage sustainable usage.
The primary goal of this project is to design and implement advanced features that help users craft more effective prompts while minimizing redundant interactions. Key objectives include:
- Introducing reusable, editable, and auto-completing prompt templates to support accurate and complete prompt formulation.
- Reducing back-and-forth interactions.
- Expanding the sustainability dashboard with detailed analytics.
- Exploring ways to raise awareness of the resource impact of prompting habits, such as integrating nudges or feedback mechanisms.
The students will address the following central questions:
- Prompt Design Optimization: What features (e.g., prompt templates, suggestions, auto-fixes) can help users reduce the number of interactions while maintaining response quality?
- User Guidance & Feedback: How can the chatbot interface guide users toward more sustainable and efficient usage patterns without limiting creativity or expressiveness?
- Visualization & Insights: What additional information should the dashboard display to provide users with meaningful insights into their prompting behavior and its impact?
- Behavioral Impact Evaluation: How can the effectiveness of these features be measured in terms of prompt reduction, user satisfaction, and sustainability awareness?
Technologies
- Frontend: Angular
- Backend: Python
- LLM API Integration (OpenAI, HuggingFace, or locally hosted models)
- Visualization: Plotly, D3.js, or Chart.js for analytics dashboard