In this guide, we explain how you can use generative artificial intelligence (AI) to write better guides for activists. We show you what prompts to use and warn for its limitations.

Why we need humans #

This article explains how AI can be used as a tool for humans who want to write better guides for activists. In a relatively short time, AI has proven to be capable to do things we previously thought could never be done by computers. At Activist Handbook, we use AI a lot.

However, as AI develops, we will continue to find new shortcomings. Below are some key shortcomings that are currently present in AI tools.

Physical human experience #

Large language models, and even multimodel models, are trained on data that is available on the web. They cannot feel, they can only predict the most logical response. It works the same way computers do not understand the weather, yet they can predict whether it will be raining based on historical data.

This means that the capabilities of AI models are limited by the data that they were trained on. Not all of the human experience has been documented into data that computers can read. For example, an AI model can only understand how emotions feel physically insofar these have been accurately documented.

When humans are put into new contexts, an AI can only guess what that experience must be like. Importantly, as AI models develop, it will become very hard to tell that AI models does not have the ability to feel the physical human experience, as they will be able to confidently pretend like they do. However, this distinction is important, because it tells us that the output of an AI might be similar to that of a human, the inner workings are not.

A free will #

An AI works the way its developers have designed it to. Even though an AI may do things the developers did not intend, they are alway directly the result of the input given by its developers.

A human is not moulded in the same way. Of course, we try to educate children, but we cannot control their entire human experience. And through our human experience, over time, our personalities and free will are formed.

As stated before, AI models only base their output on the input that its developers have decided it should and should not be trained on. This gives a lot of power to the designers of AI models.

An AI model can pretend to have a free will, but only if its developers designed it that way.

Translation versus localisation #

The introduction of large language models has dramatically brought down costs to translate text into different languages. For many purposes, especially in the case of literal translation, AI is already good enough to replace human translators. It will only get better.

This has huge consequences for humans working in the translation industry. Many translators will lose their jobs and with that livelihoods.

Assume you will soon be able to read any text, regardless of what language it is written in. Locally-run in-browser translation will become good and fast enough that they will be turned on by default. Humans will be able to provide little additional value to machine-generated literal translations.

However, humans will likely continue to be able to provide value by providing local context (for example, cultural differences). AIs are only as good as the data they were trained on, and fact is, not all local contexts are equally represented on the internet.

Marginalised perspectives #

The same goes for marginalised perspectives: communities that are not well-represented on the internet will not be understood well by AI model.

Recommended prompts #

Introduction #

To write an introduction for your article that is SEO optimized, use the following prompt (use your main keyword as the topic, and optionally add up to two other keywords as shown below):

Write a very brief introduction for a activist guide about how to [TOPIC]. Explain what people will learn in the guide. Make sure to include the following keywords: [KEYWORD 1], [KEYWORD 2]

Outline #

To determine what should be discussed in your article, and how to organise that content, brainstorm using the following prompts:

Create an outline for an activist guide about how to [TOPIC].

List questions activists may have about how to [TOPIC].

List challenges activists may face when trying to [TOPIC].

Content #

After you have created an outline, use these follow-up prompts:

Elaborate on each of the items in this outline of a guide on how to [TOPIC]:

• [OUTLINE LIST]

Answer each of the following questions for our activist guide on how to [TOPIC]:

• [QUESTION LIST]

Provide possible solutions for each of these challenges that activists face when trying to [TOPIC]:

• [CHALLENGES LIST]

Summary #

The prompt below only works in the Edge chat feature. Make sure you enable access to your page contents first, and restart the browser before using:

Write a very brief summary for this page with the main takeaways.

Local context #

List challenges that arise from differing local contexts across countries for activists trying to [TOPIC]. Provide solutions to each of those local challenges.

Getting creative #

Make a funny quiz #

Create a funny quiz for activists that have just read a guide on how to [TOPIC]. Base the quiz on a fictional story.

Prepare a workshop #

Write an outline for an interactive and kinesthetic workshop that teaches the participating activists how to [TOPIC]. Encourage peer-to-peer learning and use icebreakers and energizers.