In exploring the concept of developing Open Educational Resources with AI assistance, I queried Claude 3.5 Sonnet with the following prompt:
User:
Let’s consider copyright implications– or perhaps, 80% copyright, 20% “academic integrity.” Assuming I make a proper project out of this and publish an OER, we ought to have some sort of boilerplate text explaining the provenance of these materials.
I think overall that we are in a transitional period in which these issues have not been worked out, either legally or in terms of cultural “academic integrity” issues.
Let’s suppose:
- I want to use CC BY-SA 4.0 licensing
- I want to claim “primary author” role, but want to explicitly acknowledge the assistance of one or more LLMs
- I am not averse to naming Claude 3.5 Sonnet by name, but that I don’t necessarily want to paint myself into a corner of needing to cite every generative AI tool I might use over the course of an entire project
From that starting point, I want you to meditate on these (and any related) issues and present me with multiple scenarios for attribution/licensing, as many as seem pertinent for an intelligent discussion on these issues.
Claude 3.5 Sonnet:
This is an important and complex issue that deserves careful consideration. You’re right that we’re in a transitional period where norms and best practices around AI attribution in academic and creative works are still evolving. Let me present a few potential scenarios and approaches for attribution and licensing, along with some pros and cons to consider.
AI Attribution Scenarios for OER Materials
Scenario 1: Generic AI Acknowledgment
Boilerplate Text:
“This work was created by [Your Name] with assistance from artificial intelligence language models. It is licensed under CC BY-SA 4.0.”
Pros:
- Simple and straightforward
- Doesn’t commit to naming specific AI tools
- Allows flexibility for using multiple AI tools throughout the project
Cons:
- Lacks specificity, which may be seen as less transparent
- Doesn’t distinguish between different levels of AI involvement in various parts of the work
Scenario 2: Specific AI Acknowledgment
Boilerplate Text:
“This work was primarily authored by [Your Name] with assistance from the Claude 3.5 Sonnet AI language model developed by Anthropic. Additional AI tools may have been used in the creation of this work. It is licensed under CC BY-SA 4.0.”
Pros:
- More transparent about the specific AI tool used
- Still allows for the use of other AI tools without needing to list them all
- Clearly establishes your role as primary author
Cons:
- May need updating if you switch to primarily using a different AI tool
- Could be seen as promoting a specific AI product
Scenario 3: Detailed Attribution with Percentages
Boilerplate Text:
“This work is primarily (80%) authored by [Your Name], with significant assistance (20%) from artificial intelligence language models, including but not limited to Claude 3.5 Sonnet by Anthropic. The exact percentage and nature of AI involvement may vary between sections. This work is licensed under CC BY-SA 4.0.”
Pros:
- Very transparent about the level of AI involvement
- Provides a clear picture of your contribution versus AI assistance
- Allows for variation in AI usage across different parts of the work
Cons:
- Percentages may be difficult to accurately determine
- Could raise questions about how the percentages were calculated
- May complicate the perception of authorship
Scenario 4: Section-Specific Attribution
Boilerplate Text:
“This work is primarily authored by [Your Name] and is licensed under CC BY-SA 4.0. Sections that received significant assistance from artificial intelligence language models are marked with an asterisk (*). The primary AI tool used in this project was Claude 3.5 Sonnet by Anthropic, though other AI tools may have been used to a lesser extent.”
Pros:
- Allows for more granular attribution
- Clearly distinguishes between AI-assisted and non-AI-assisted sections
- Provides transparency without needing to list every AI tool used
Cons:
- Requires ongoing management to mark AI-assisted sections
- May lead to questions about unmarked sections and the extent of AI use
Scenario 5: Generative AI as a Tool
Boilerplate Text:
“This work was created by [Your Name] using various digital tools, including word processors, research databases, and generative AI language models. Generative AI was used as a drafting and ideation tool, with all output reviewed, edited, and approved by the human author. This work is licensed under CC BY-SA 4.0.”
Pros:
- Frames AI as one of many tools used in the creative process
- Emphasizes human oversight and decision-making
- Avoids the need to specify particular AI tools
Cons:
- May be seen as downplaying the role of AI
- Doesn’t provide specifics about the extent of AI usage
Considerations for Choosing an Attribution Approach
-
Transparency vs. Simplicity: More detailed attribution provides greater transparency but can complicate the licensing and attribution statement.
-
Flexibility: Consider how your AI usage might change over the course of the project and choose an approach that allows for that flexibility.
-
Academic Integrity: Think about how your approach aligns with current academic norms and emerging best practices for AI disclosure in scholarly work.
-
Legal Considerations: While the legal landscape is still evolving, consider how your approach might be viewed in terms of copyright and authorship rights.
-
Audience Perception: Consider how your target audience (e.g., students, fellow educators) might perceive different levels of AI involvement and attribution.
-
Practicality: Think about the practicality of maintaining your chosen attribution approach throughout the project and in future updates.
-
Institutional Policies: If applicable, consider any guidelines or policies your institution might have regarding AI use in educational materials.
Remember, as the field evolves, best practices for AI attribution are likely to change. It may be worthwhile to include a note that your attribution approach is based on current best practices and may be updated as norms evolve.
I’ve outlined five different scenarios for attribution and licensing, each with its own set of pros and cons. These scenarios range from generic AI acknowledgment to more detailed, section-specific attribution. I’ve also included some key considerations to help guide your decision-making process.
This is a complex issue with no one-size-fits-all solution. The best approach may depend on factors such as your specific project goals, your audience, and the norms in your field. It might also be worth consulting with colleagues or legal experts in open educational resources for their perspectives.
Would you like to discuss any of these scenarios in more detail? Or perhaps explore how these attribution approaches might be practically implemented in your OER project?