November 15, 2024, by Brigitte Nerlich

Chatting with chatbots about the climate crisis

Last week, I had another adventure in AI land. This started by accident, as so many adventures do. It all came about because I read about an interesting symposium organised in Amsterdam by Anaïs Augé’s and Gudrun Reijnierse on “Public responses to the language of science communication: Uptake, acceptance, resistance”. As part of this symposium, Anaïs arranged a workshop on “Imaginaries of the climate crisis on ChatGPT”. That tickled my curiosity.

Chatbots and the climate crisis

I went to ChatGPT and asked outright: “What do you make of the climate crisis?”. The answer came back instantly including in bold letters: “The science is clear”. I hadn’t expected such as forthright answer. The rest was equally in line with what is generally called consensus science. I then tried the question out on Claude, Perplexity, Notion and Gemini and got similar answers (these transcripts still need to be analysed). (I only used LLMs that were directly accessible to ordinary mortals)

Like ChatGPT, Claude said quite clearly: “There is a global scientific and political consensus that rapid decarbonization of the global economy is essential to avoid the worst-case scenarios of climate change.” All the chatbots I asked told similar stories about the science and listed urgent actions that could/should be taken to mitigate climate change impacts.

Unlike the other bots though, Claude turned around at the end of its* answer and told me: “I would encourage you to stay informed on the latest climate science and policy developments from authoritative sources. And I’m happy to discuss further ways that individuals, communities, and governments can take action to address this critical issue.”

Chatting with Claude about the climate crisis

I then started a conversation with Claude during which I needled it a bit with some climate scepticism. This astonished Claude who asked: “I notice you’re questioning the certainty of the science – what specific aspects make you uncertain?”. I then admitted that I was just testing out various chatbots on how they talked about climate change, which made it sort of ‘chuckle’.

We then went on to ‘discuss’ issues around climate change, climate denial, the pros and cons of AI in the context of climate change, AI safety, responsible AI, the power of corporations and politicians etc. and how one could mitigate not only the impacts of climate change but also the impacts of AI on climate communication and the climate itself.

Affordances

In that context, I used the concept of ‘affordances’ by saying “One more thing might be of interest: J.J. Gibson once introduced the concept of affordances which refers to what the environment offers an individual (e.g. a door handle ‘affords’ grasping). I was once asked how one can improve climate change communication in order to change people’s behaviour. I said: ‘You can communicate until the cows come home. Unless people live in an environment where they can actually do things differently, they won’t do things differently.’” (And it’s also important to note that these ‘things’ should be easier to achieve and more attractive than their alternatives, see here).

This quite ‘intrigued’ Claude and it then gave me hints about how and with what academic tools I could pursue that topic further: “This relates to what sociologists sometimes call ‘social practice theory’ – our behaviors are shaped by material infrastructure, social norms, and institutional arrangements more than by individual choice alone. The focus on individual behavior change through communication can sometimes distract from these deeper structural issues.” It also thought that “the intersection of environmental psychology, political ecology, and climate communication” was worth exploring. Interesting.

Podcast

Then I realised that I didn’t have time to analyse my conversation with Claude in any depth (I had visitors and needed to engage in some actual conversations). So, I fed the whole conversation straight into NotebookLM which turned it into a Deep Dive podcast I could listen to on the go. Again, I was fascinated by what happened during this meta-level conversation.

The female voice was ‘me’ telling the male voice about her/my conversation about the climate crisis with an AI (they didn’t mention Claude). And just as Claude had gone beyond my rather messy prompts, so the podcast prompted by my conversation with Claude also went well beyond what Claude and I had talked about.

I was pleased that they absolutely ‘loved’ my introduction of the term ‘affordances’. I had talked about this at a 2008 Oxford climate communication conference, but nobody had taken much notice … I had to wait 16 years for some bots to chat with me about that!**

The end of the podcast was intriguing, as it was much more focused on inspiring action than I had expected. If you want to listen to the podcast, here it is: Deep Dive ‘podcast’ about the climate crisis based on my conversation with Claude.

*What are Claude’s pronouns?

**But of course, as I discovered early on, all these ‘AIs’ seem to have an in-built flattering or ingratiating or sycophantic function which makes them quite dangerous.

Image: Rawpixel

• Share on Facebook
• Share on Twitter
• Share on Google+
• Share on LinkedIn