Here I attempt to rank and summarise the most pertinent principles (and values) in today’s AI Ethics.

<aside> <img src="https://prod-files-secure.s3.us-west-2.amazonaws.com/5bf8f97c-b948-4af2-b064-edfc74b18e1f/f0d2d963-fd9a-42ea-ac2b-1cbd714621bb/BBX_Heart-Box-Gold-SmlBG-280w.png" alt="https://prod-files-secure.s3.us-west-2.amazonaws.com/5bf8f97c-b948-4af2-b064-edfc74b18e1f/f0d2d963-fd9a-42ea-ac2b-1cbd714621bb/BBX_Heart-Box-Gold-SmlBG-280w.png" width="40px" /> This work is licensed under CC BY-NC-SA 4.0

</aside>

Introduction

We still need “principles to practice”

Going into 2024, the world of ethical AI (and Tech Ethics more generally) still suffers from a glut of guidelines and a dearth of solutions. In other words, it’s all fart and no shit.

Go take a look. It’s not hard to find the conceptual, philosophical groundwork on which AI Ethics is being built – lists of values, principles, frameworks, etc. – but I frequently hear people calling for practical tools and processes for how to actually apply it. There is still a great need for resources to help us apply moral philosophy to the actual ones-and-zeros of building and managing digital technologies.

Nevertheless, I hope the list below will be of use in your AI Ethics efforts, as a kind of building block on which to stack more detailed, specific and applicable work.

The stack sorta kinda maybe looks something like this. Thanks to Adobe Firefly for the sci-fi-fantasy pyramid image, and textstudio.com for the 3D font.

A starting point for your ethical AI policy

It can be hard to know where to set off when designing policy and methods for ethical technology. One of the most fundamental (and, I think, accessible) layers in the edifice is principles and values. The idea is that you can select those principles/values that are most important or relevant to your context of operation, prioritise them, then use this refined list as a framework on which to develop your processes, rules, products etc.

This process is at the heart of Value-based Engineering, which I learned about from a master, Sarah Spiekermann, while collaborating with her and others on the new global standard based on that process: IEEE 7000. Here’s a paper outlining some of the concepts: Value-based Engineering for Ethics by Design.

<aside> ⚠️ But are these “principles” or “values”? Either. Both. Take your pick. There are semantic differences between principles, values and other similar words, but for the scope of this (high-level) project, I’m lumping them together.

</aside>

Ranked: Top Principles for Ethical AI

Okay, let’s get to the meat of this particular sandwich.

I gathered a bunch of broad-reaching and reputable lists of ethics principles, as they apply to modern technology, particularly AI. Some were standalone lists, while others were meta-analyses of many other underlying lists. Then I roughly (and somewhat unscientifically) condensed them into a ranked list of the top principles.

This got me to a list of 103 principles. I then tried my best to group them into common terms. I had to use a little creative licence but tried to maintain the spirit of each principle while also shortening the overall set. This was the result:

About the source data

<aside> ℹ️ Here’s my source data, including links to the publications: ‣

</aside>

Notably, all five of the meta-analyses I found were also covered in this article: The Proliferation of AI Ethics Principles: What’s Next?, from Montreal AI Ethics Institute (2021), so it deserves a mention.