Articles, English

An AI-powered chatbot trained to discuss climate change

By Mitchell Dickau


Reading time - Temps de lecture: 3 minutes

Using the power of AI to help reduce the gap between our best understanding of the science behind climate change and public perceptions of our changing climate.

Polling from 2021 shows that 7% of Canadians believe that there is “little evidence” or “no evidence” of global temperature rising and 24% of Canadians believe there is “some but not conclusive evidence”. In contrast, the world’s leading climate scientists and all the 197 member nations of the United Nations Framework Convention on Climate Change (UNFCCC) approved the following statement in the Intergovernmental Panel on Climate Change’s (IPCC) latest report:

It is unequivocal that human influence has warmed the atmosphere, ocean and land. Widespread and rapid changes in the atmosphere, ocean, cryosphere and biosphere have occurred.

Furthermore, the same poll from 2021 showed that, of respondents who believed that the climate is warming, only 75% believed that observed warming is caused by humans even though the IPCC attributes nearly 100% of observed warming to human activities. 

The disconnect between the scientific understanding of Earth’s climate system and the public perception of climate change demonstrates the need for science communication and climate change education. Together with Reimagine AI – a Montréal based artificial intelligence (AI) creative studio – me and several other LEADS interns began to create an AI-powered chatbot that can discuss climate change and climate change related issues with users. Our objectives were to develop a chatbot that makes learning about climate change interactive, easy, fun, and accessible to people with a wide range of ages and educational backgrounds.  Reimagine AI has developed a platform known as SECONDSOUL that allows them to build interactive AI-powered minds. One example of their work is a character they have developed that is meant to imitate Albert Einstein. While I have no expertise in AI, my expertise in climate science and the dynamics of the climate system made me well-equipped to help on the scientific side of our project.

User speaking to Albert Einstein chatbot created by Reimagine AI. Source: Reimagine AI website.

Figure 1 – User speaking to Albert Einstein chatbot created by Reimagine AI. Source: Reimagine AI website.

My work included compiling sets of common climate change related questions, such as questions to do with climate mitigation, physical phenomena causing climate change, observed changes in the climate system, and climate impacts. Our aim was to include questions that would allow for users with a wide range of prior knowledge related to climate change to be able to learn from their discussion with the chatbot, whether that be a user with no prior knowledge on climate change or a user with strong background in climate change. 

Once we had developed hundreds of questions, I began to shift my focus to creating several succinct answers (about the size of a tweet) to each question. The length constraint of the answers was designed to ensure that the user experience is easy and fun. However, for more technical questions related to the physical science of climate change, the length constraint of answers posed challenges. To account for this issue, many of the answers ended with prompts that encourage the user to ask follow up questions.

This climate change chatbot by Reimagine AI is still in the early stages of development. As the project moves forward, more climate change related questions and answers will be compiled and, eventually, these questions and answers will be used along with other data to train the AI chatbot. Once the AI chatbot is released, it appears as though it will be a helpful tool for climate change education.

Keywords: Artificial intelligence, science communication, climate change education

Headshot of Mitchell Dickau

Mitchell Dickau is a PhD student in the Matthews Climate Lab at Concordia University and a member of LEADS since 2020. His research focuses on questions surrounding the concept of a remaining carbon budget which describes the remaining amount of CO2 that humans can emit to the atmosphere before a given temperature threshold is met.