A Framework for Classifying AI-Generated Content
Levels of Human Involvement and Human Verification
The proliferation of generative AI has created a new and complex information landscape. Content can now be created by humans, by AI, or by some combination of the both, yet the term "AI-generated" is often used as a simple, catch-all label. This lack of nuance makes it difficult to assess the value, reliability, and authorship of a piece of work.
To address this ambiguity, we propose a two-dimensional framework for categorizing content. This system provides a more precise lens for understanding and managing AI-assisted or AI-generated outputs. The two dimensions are: Levels of Human Involvement and Human Verification Status
Dimension 1: Levels of Human Involvement
This dimension measures the degree to which a human is involved in the creative process, divided into five distinct levels.
1. Full Human Involvement
This is content created entirely without generative AI assistance. The ideation, drafting, and final execution are all performed by a human.
Example: An paper researched and written from scratch by an academic; a novel penned by an author.
2. High Human Involvement
In this category, AI acts as a sophisticated tool to augment or transform human-created work. While a machine is involved, the core intellectual property and authorship remain fundamentally human.
Example: A human-written report translated into another language by an AI. The original meaning and substance are human-authored.
3. Medium Human Involvement
This level represents non-trival contributing by both human and AI. The human guides the process by providing detailed instructions, source material, or by heavily editing and restructuring an AI-generated draft.
Example: An AI generated podcast on a research paper with the aim to introduce the research to a more general audience.
4. Low Human Involvement
This includes content generated from a simple human prompt, where the AI performs the vast majority of the creative and structural work. Human contribution is limited to the initial idea or command.
Example: Asking an image generator to Ghiblify a photo.
5. No Human Involvement
This is content produced by an autonomous AI system without a specific human prompt for outputs.
Example: An AI that automatically generates hourly summaries based on the latest news.
Dimension 2: Human Verification Status
Independent of the level of human involvement in its creation, any piece of content can be classified as either verified or not. This binary classification—Human-Verified or Not Human-Verified—acts as a crucial layer of quality control and trust.
Human-Verified: This label indicates that a human with appropriate expertise has reviewed the final output for accuracy, quality, and adherence to standards before its release. This "human-in-the-loop" step provides a critical assurance of reliability.
Not Human-Verified: This indicates content that has been published directly from an AI system without a final human review.
For instance, a "Low Involvement" article (e.g., a sports game summary) could be published instantly as Not Human-Verified, or it could be sent to an editor first and be published as Human-Verified.
Applications and Importance
Adopting this two-dimensional framework offers a powerful way to manage our new information ecosystem.
Transparency and Trust: It provides consumers with a clear understanding of how a piece of content was made, allowing them to better judge its credibility. We should demand stricter controls and clear labeling for content that is both low in human involvement and not human-verified.
Attribution and Value: The levels of involvement provides a nuanced way to recognize and credit the human effort invested in a work. This is crucial for navigating emerging questions around copyright, authorship, and compensation.
Ethical and Editorial Standards: Organizations can use this framework to establish clear internal policies for using AI, ensuring quality control and accountability for their published output.
By moving beyond the simple "AI-generated" label, this framework fosters a more responsible, transparent, and trustworthy integration of artificial intelligence into our world.
