Hopefully this addresses such questions

bsky.app/profile/dral...

www.swp-berlin.org/publications...

Ultimately, the IPCC should develop an AI policy robust to multiple futures. As a model institution, its choices will shape how other fields approach scientific assessment in the AI age.

13/13

Working paper: www.swp-berlin.org/publications...

Options for the IPCC: A thoughtful middle path. Leverage AI to enhance equity (e.g., overcoming language barriers) whilst carefully weighing risks of official AI tools that could undermine the stable, shared foundation of facts needed for international cooperation.

11/13

3) Anticipatory resistance: Scientific communities adopt precautionary approach; restrictive AI policies risk perpetuating epistemic inequities

4) Public backlash: AI adoption becomes stigmatised; any AI use threatens IPCC legitimacy

10/13

1) Rise of the agents: Widespread agentic AI transforms workflows; IPCC must optimise outputs for both human and AI readers

2) Superior truth machine: AI viewed by some as more objective than humans; IPCC faces competing AI-generated assessments

9/13

Four plausible futures (thought experiments):

We explore scenarios informed by two axes: societal/scientific acceptance (trusted ↔️ taboo) and degree of automation in knowledge processing (human-driven ↔️ AI-driven).

8/13

💬 Communication: Should the IPCC create an official chatbot to pre-empt third-party interpretations that fragment carefully built consensus?

7/13

✍️ Literature assessment: How do we maintain human deliberative consensus when AI-generated 'shadow assessments' could contest IPCC findings using the same literature?

6/13

Three functions, multiple challenges:
📚 Literature identification: Can AI agents comprehensively scan tens of thousands of papers whilst preserving inclusion of grey literature and Indigenous Knowledge?

5/13

AI disrupts not just report-writing workflows, but this deeper social function of building consensus, both scientific and political, around climate reality.

4/13

Why this matters: The IPCC isn't merely a literature synthesis machine. It creates a scientifically authoritative but politically legitimised consensus that functions as 'de facto governance', shaping research priorities and policy options.

3/13

The context: The seventh IPCC assessment cycle (AR7, ending 2029/30) will be the first conducted in an era of widespread generative #AI. This raises critical questions beyond technical efficiency.

2/13