Communicating AI governance proposals to non-experts

The current developments in AI governance have brought the topic into the center of public discourse. This is a chance for organizations that have been developing sensible AI policy proposals to present them. 

However, when talking to people who are new to the topic, I keep noticing that certain kinds of material are either missing, or possibly exist somewhere but are hard to find. 

A list of material that seems missing

• An explainer of AI risk in five bullet points. The fourth should be a one-sentence explanation of policy demands, the last should contain a link to a website with explanations in different degrees of length, depth and technical experience.
• A flyer that you can print and hand to people, not larger than an A5-sized sheet of paper.
• An AI reading list for policymakers, again in different degrees of length, depth and technical experience.
• A website giving an overview of AI-risk organizations, addressed at different kinds of readers. (Existing websites that I know seem to be aimed at people already working in AI safety or at least already wanting to do so.)
• A list of experts, speakers, panelists, and interview partners as contacts for policymakers and journalists.
• A list of (good) newspaper articles and nontechnical blog articles on AI risk.

Some remarks on the form of the material

Language

• If you try to reach people outside the English-speaking world, you will likely lose part of your audience if you write and speak in English.
• Even fluent or 'business fluent' English speakers often engage less with English text than with their native language.
• Therefore, it is important to have translations of the above into the locally relevant language. This may be a very low-hanging fruit.

Institutions and context

• Most people know less about political institutions than you do. Therefore, general information will usually suffice for a flyer aimed at the public. However, when addressing political decision-makers, you should take into account their competencies, responsibilities, and interests—as well as the interests of their respective institutions.
• If your text is copied & pasted and refers to the context of another country, people will conclude that your text is not for them.

Rhetorics and reputation

• If you send a text or hand a flyer to someone, it is helpful if the person can share it with other people. If you write something that can be perceived as insulting to someone else, people may be repelled and not be willing to pass it along.
• Be wary of guilt by association. Others may voice demands for AI governance that are bad, or seem bad out of context. To non-experts, these demands can seem similar to yours – which can make communication more difficult.

Presentation on websites

• Make sure that you know whom the website is meant to address. It is totally fine to put something as clear as 'an AI explainer for everybody' or 'Priorities for European AI policy' in your website's subtitle.
• Testing how your website is perceived is helpful and not very difficult. Just ask people (or an LLM) to give you feedback.

Finally, back to concepts

Assuming good concepts and proposals already exist, presentation form currently has priority. That said, content-related demands remain relevant too. The first things that people see from you should be much more concrete than "We want better AI governance."

You should also be prepared to address confusion. For example, you^[1] should have a clear answer to whether Europe should build more datacenters and cut regulation – and be able to explain how this fits with calls for a slowdown or pause, given that the two can seem to pull in opposite directions.

Thanks to @Flo 🔸  for remarks on the draft.

1. ^{^}

   In this case: you = people active in European governance discussions.