Originally published at the #semanticClimate blog ↗
Chatbots serve as a useful tool for knowledge, answering questions within seconds. However, when the topic is about something as significant as climate change, accuracy and reliability matters as much as speed.
Motivated by this need of reliable information, we introduce the Climate Academy chatbot, which is a collaborative effort by Climate Academy, semanticClimate and Open Knowledge Foundation (OKFN), as a part of OKFN’s AI Learning Labs initiative. The goal was to make climate knowledge more accessible to everyone, keeping information clear, factual, and credible.
Objectives
- Semantic Search: Understand meaning, not just keywords
- Retrieve Relevant Content: Find the most relevant sections from the Climate Academy book.
- Grounded Answers: Generate responses strictly from retrieved content — no hallucinations.
- Source Citations: Every answer includes a reference to the original book section
- Multi-User Support: Handle concurrent queries with a production-ready backend
- Production-Ready: Built for reliability, scalability, and real-world deployment
The Knowledge Behind the Chatbot
The Climate Academy chatbot is based on Climate Academy: The Student Guide, written by Matthew Pye. Written for high school students, it introduces climate science in a way that is scientifically accurate and easy to understand.
How the Climate Academy Chatbot Works
Climate knowledge is often difficult to navigate as it is scattered across a plethora of long reports. The Climate Academy chatbot makes learning easier by providing clear and reliable answers in simple language.
- Knowledge resources (PDF/HTML)
- Section Parser
- Text Chunking
- Embedding
- Vector Store
- User Query
- Retriever and LLM
- Answers and Citations
- Retriever and LLM
- User Query
- Vector Store
- Embedding
- Text Chunking
- Section Parser
What Makes a Climate Chatbot Responsible
A chatbot is responsible when it is built on trusted sources, clear design choices and thorough testing, so that it can answer questions appropriately. The following factors describe what makes a chatbot responsible:
- Reliable Knowledge: The chatbot uses trusted climate resources to ensure accuracy and reduce misinformation.
- Transparency: The chatbot should show the sources of information from where they have constructed the answer, in order to verify its credibility to the users.
- Fairness: The chatbot should be able to express how climate change affects different communities from different geographical regions in a balanced and respectful way.
- Accessibility: The chatbot uses simple language to present climate knowledge in such a way that everyone understands.
Conclusion
The Climate Academy Chatbot provides reliable information from Climate Academy: The Student Guide in clear and accessible way for an interactive learning experience.
About
Open Knowledge’s AI Learning Labs is an initiative that aims to experiment with AI, translate knowledge from social sector organisations around the world, and produce public, multilingual AI-literacy resources tailored for organisations addressing similar issues elsewhere.
Together, we will catalyse learning and develop replicable methods to help organisations build AI skills, use AI responsibly, and develop their own AI projects. All resources will be openly available at School of Data.
Join the conversation:
- Open Knowledge Forum
- School of Data community
- Contact our team: info@okfn.org
This project has been made possible thanks to the generous support of the Patrick J. McGovern Foundation (PJMF). We are grateful for our ongoing partnership in promoting digital literacy and investing in AI for the public good. Learn more about its funding programmes here.
Facts Only
* Chatbots serve as a tool for knowledge, answering questions quickly.
* The Climate Academy chatbot was developed by Climate Academy, semanticClimate, and Open Knowledge Foundation (OKFN).
* Objectives included semantic search to understand meaning, not just keywords.
* The system retrieves relevant sections from the Climate Academy book.
* Answers are generated strictly from retrieved content with no hallucinations.
* Every answer includes a reference to the original book section.
* The system supports multi-user queries with a production-ready backend.
* The chatbot is based on Climate Academy: The Student Guide by Matthew Pye.
* The knowledge workflow involves knowledge resources, section parsing, text chunking, embedding, vector store, user query, retriever, and LLM.
Executive Summary
The Climate Academy chatbot is designed to make climate knowledge accessible by synthesizing information from the "Climate Academy: The Student Guide" into clear, factual, and credible answers. It functions by employing semantic search to understand meaning beyond keywords, retrieving relevant content sections from the source material, generating answers strictly grounded in that retrieved content to prevent hallucinations, and providing source citations for every response. The system is built upon a knowledge retrieval pipeline involving text chunking, embedding, vector storage, section parsing, and interaction with a Large Language Model (LLM) to handle concurrent user queries in a production-ready manner.
The responsible design of the chatbot hinges on four principles: relying on reliable knowledge from trusted sources, ensuring transparency by citing information origins, promoting fairness by presenting impacts across different communities neutrally, and enhancing accessibility through simple language. Ultimately, the tool aims to provide an interactive learning experience based on established climate science while maintaining accuracy and accountability.
Full Take
The architecture reveals a commitment to grounding AI in explicit, structured knowledge rather than pure generative ability. The process—moving from keyword search to semantic understanding, retrieving specific chunks, and enforcing citation—is a direct attempt to mitigate the inherent risks of LLM-based information synthesis. This places the system squarely in the Constructive Mode, focusing on educational utility rooted in verifiable texts.
The focus on 'Fairness' and 'Accessibility' as core responsibilities moves beyond mere technical accuracy into ethical AI deployment concerning complex topics like climate change. The pattern suggests a reaction against opaque knowledge delivery, where the mechanism itself must be transparent to build user trust. A critical implication is whether simplifying complex scientific realities (like community-specific impacts) inherently risks oversimplification or erasure if not handled with careful contextual layering. The ongoing initiative by OKFN highlights a broader systemic effort to decentralize and democratize AI learning resources, suggesting that reliability in public knowledge systems requires collaborative oversight extending beyond technical implementation to encompass ethical representation.
What steps are missing in validating the 'Fairness' objective? How can accountability be mathematically modeled when dealing with subjective concepts like "respectful expression" of disparate community impacts? If the goal is true accessibility, does reliance on a single source, even a trusted one, risk inadvertently imposing a singular, authoritative narrative over necessary, nuanced regional perspectives?