Giving credit where it's due: transparency in creating with AI

Recently I’ve been thinking a lot about the nature of creativity in general, and especially when working with AI as a genuine creative partner.

Many people today will view the very sentence above with horror, I know. Many others are happily creating all kinds of things with AI. I’m not going to go into all the positions and issues around AI and creativity in this post, but I believe all those viewpoints are personal and completely valid.

Right now I want to highlight that a lot of the friction around creative works (I REFUSE to use the word content ¹ ) involving AI is because the AI’s involvement is often “hidden” in the work. Students using AI to generate things for assessment, deep fakes, AI-generated science, social media "botshit", etc all cause problems essentially because some people might not realise they were generated by AI.

I believe this is hugely urgent to solve in a global way. We need to understand the provenance of our digital diets as much as our physical ones. There are many reasons for this, not least of which is the wide variety of opinions about the simple validity of AI output I mentioned above, but it also helps solve misinformation and abuse.

I see three levels to transparency:

1. "This used AI"

The very first most basic thing we can do is to be transparent and declare that AI was involved. Until now this has taken the form of a “sparkle icon” or a tag. Moodle adds this with anything it generates. Many image generators add a watermark. The C2PA is an emerging standard for embedding Content Credentials as metadata in images, videos and pdfs. Safe AI proposes a simple icon.

Unfortunately take-up of even this basic level of transparency is low. The dominant culture among those using AI is to communicate nothing at all about AI involvement. Large platforms for creative works such as Spotify and Youtube do not ask uploaders even this simple question, and there is no way to filter searches to avoid (or discover) AI works.

2. "Levels of AI assistance"

A lot of things are hitting the media as “created by AI” when that’s almost never entirely true - a person initiated that creation, and their level of effort in creating the final work may range from trivial (a single prompt) to many hours of editing, steering, and packaging. Effort is a popular metric that people often use to help decide how impressed they are by a creative work.

An emerging approach is to indicate how much AI is involved (or alternatively, how much human effort was involved). This simplifies the involvement of AI into a scale that paints a spectrum ranging from “No AI used at all” to “Completely AI-generated”. A few people have proposed some scales like this but they are not in wide use yet, some are: IBM’s AI Attribution Toolkit, AI Assessment Scale, Icons for Human-Machine Collaboration or even a simple percentage value (eg 25% AI).

While not perfect, I do believe this is a very useful shorthand way to communicate about these issues as it’s very easy to understand. It can be used verbally and for real-world works, as well when distributing creative works online. It would be great for internet platforms to add support for scales like this, as it would encourage more people to specify the level for any works they place online.

The problem with this approach is that we still can’t tell what role AI played in the creation, and that’s sometimes important. Which leads us to:

3. “Roles of AI”

Most serious creative works in history (music recordings, novels, movies etc) are not the work of a single person, or have many different components, and we have many existing methods of crediting individuals for their roles in the finished product. A single movie might have thousands of contributors, and their roles are all recorded in the credits.

I believe this approach gives us a well-understood and ethical way to credit AI in particular projects.

For example, I have been a musician and producer for many years, mostly for my own pleasure. In recent years I’ve been experimenting more and more with tools like Suno.ai and Udio.com to help produce music, with great results. It is possible to use these tools in a very low-effort way (“Make me a song about how ice-cream is nicer in summer” will give you a full song in 30 seconds) but it is also possible to get deep into custom workflows to get particular advanced results you’re aiming for (I typically am spending 5 hours or more for each song).

The credits for a typical song (these are all standard fields for any music distribution platform) might look like this:

• Artist: Mantis Audiogram
  (my project name)
• Composer: Martin Dougiamas
  (I am the source of the core creative idea, genres, intent and structure)
• Lyrics: Martin Dougiamas
  (I wrote the words)
• Band/Performer: Suno.AI
  (It’s the entity "performing" the music and singing the words, like session musicians or an orchestra).
• Producer: Martin Dougiamas and Suno.AI
  (This is a true collaboration. The human guides the sound through prompts, and the AI makes production choices to realise that sound)
• Engineer: Suno.AI and Mixea.AI.
  (They handled the technical aspects of mixing and mastering the final track)
• Publisher: Martin Dougiamas
  (I own the copyright and I am managing its distribution)
  

For a book, you might have:

• Author: Human. You provide the core plot, the themes, the narrative structure, and the final voice. The AI is a powerful drafting tool, but you are the author.
• Researcher Editor: Human & AI. The AI can collate information or perform a first-pass copy edit with incredible speed, but the human is essential for fact-checking, developmental editing, and ensuring stylistic cohesion.
• Illustrator / Cover Designer: Human & AI. The human provides the artistic direction and concept, while the AI can generate stunning visual options to be curated and refined.
• Publisher: Human
  

For a movie generated with something like Google VEO 3, you might have:

• Director & Writer: The Human
  (These roles become even more critical, responsible for the overarching vision, story, and the meticulous prompting that guides the entire production)
• Editor: The Human
  (Assembling the AI-generated shots into a coherent narrative, controlling pace and rhythm, is still usually a human art)
• Actor, Cinematographer, Production Designer: The AI.
  (These traditional roles are absorbed by the generative model. The "performances," "camera work," and "set design" are all executed by the AI based on the director's explicit instructions)

I’m not seeing anyone actually doing this much in practice yet, but I hope to!

What I like about this model is its respect for human intent. In each case, the human creator remains the central, guiding force—the composer, the author, the director. It positions the AI as an incredibly powerful and versatile partner that can perform well-defined roles within a project.

(From information like this we can even guess a measure of the human effort involved. I personally wish there were a standard metadata field where we could note the time and intensity of the human’s contribution, but this has never been a thing in all of history for any creative work, so it will probably remain on forewords, or liner notes or blogs).

As we build the future of learning and share our creativity, transparency on all these fronts will be paramount. I hope you are sold on the idea of using all these techniques to provide as much transparency as you can to ALL your AI-Assisted works, even if it’s just an email.

AI is not your personal secret sauce, we ALL have access to it now, so let’s demystify the process, give credit where it's due, and foster a healthier, more honest relationship with the powerful new creative tools at our disposal.

More coming soon on creativity in future posts.

Credits:
Author: Martin Dougiamas
Brainstorming on specific roles: Gemini
Image: Imagen 4

¹ “content” is a word that comes from today’s advertising-driven mega-platforms, which see themselves as containers for “content” that provides something to attract viewers to look at advertising hung around it.