9 Generative Artificial Intelligence
BCcampus Open Education Team
Here are guidelines to consider if you plan to use generative AI tools during the OER content creation process.
As the technology in this space is evolving rapidly, these guidelines will continually be reviewed and updated. This is a living resource that will be updated as new developments unfold.
What is Generative Artificial Intelligence (AI)?
For the purposes of this document, we define generative artificial intelligence (AI) as a type of artificial intelligence that is used to create images, text, audio, video, computer code and other types of content via text prompts from a user. Generative AI can be in the form of a standalone tool or can be incorporated or integrated into other content creation tools.
Examples of generative AI tools
There are hundreds of generative AI tools with more being developed each day, so it is impossible to provide a definitive list of all the tools that are covered by these guidelines. Instead, to help define what we mean by generative AI further, we are providing some examples of the types of tools that are considered generative AI for the purposes of these guidelines.
| Tool | Output |
|---|---|
| ChatGPT | Text |
| Dall-e or midjourney | Images |
| Eleven Labs | Voice/speech |
| Make a Video | Video |
| Riffusion | Music |
| GitHub Copilot | Computer code |
While this is not an exhaustive list, it should provide you with a general idea as to the types of tools and outputs we are referring to with this set of guidelines.
Ways You Could Use Generative AI to Create OER
Here are some ways that generative AI could be used when creating or adapting OER:
- To create questions sets, case studies, and other types of instructional resources.
- To analyze a photo to create alt text for accessibility purposes.
- To create illustrations and photo-realistic images for both decorative and instructional purposes.
- To generate scripts that can be used for videos and podcasts.
- To create instructional videos.
- To generate sentences, paragraphs, and chapters for a textbook.
- To analyze and create summaries of longer sections of text.
- To automate the creation of an audio version of text, usually for accessibility purposes.
- To translate text to another language.
Intentional vs. unintentional use
With the proliferation of generative AI tools and the continual integration of these tools into other software packages, you may not know that a tool you are using to create learning materials is using generative AI. Therefore, these guidelines are intended to be interpreted with leniency and flexibility to allow for the possibility that generative AI use may not always be visible or apparent to the person who is using a tool.
Considerations and Risks
While these tools can be of great value, there are numerous ethical concerns and potential risks you should be aware of when you consider using generative AI tools to develop OER.
- Lack of transparency. The source of the training data and the programming logic used by many generative AI tools is not always made available to the public. Indeed, even the companies that develop the tools may not be able to explain exactly how they work, or how they arrived at the outputs they did.
- Bias. Because there is a lack of transparency in how the tools are constructed and what data is used to train them, this can lead to biases being present in the output.
- Accuracy. Generative AI systems can sometimes produce inaccurate or made-up answers (also referred to as hallucinations).
- Intellectual property (IP) and copyright. There are three areas where copyright and IP should be considered; the content that the AI tool has been trained with, the content that the AI tool generates, and using AI to generate summaries of copyrighted content.
- Use of content to train AI models. Many generative AI tools have been trained on copyrighted works often without the permission of copyright holders. While organizations such as Creative Commons argue that this use is considered fair use under current copyright legislation, there are a number of lawsuits where creators are arguing that AI tools are creating unauthorized derivatives. This also means that AI-generated content could be subject to copyright claims.
- Applying copyright to generated output. Legal decision and rulings around copyrighting AI generated content has been very clear in the United States. AI generated content is not human made and therefore cannot be protected by copyright. In Canada there has not been as definitive legal rulings around this, although the emerging consensus in Canada is that Canadian copyright law will follow closely with the US when it comes to copyright and AI due to the shared international copyright and trades agreements the two countries have with each other.
- Using AI to generate summaries of copyright work. It is unlikely that using generative AI to summarize copyrighted content is a violation of copyright as the AI generated summary is machine and not human generated.
- Sustainability. Generative AI uses massive amounts of electricity to operate, which has led to examinations as to how environmentally sustainable generative AI is.
Guidelines and Recommendations
- Be cautious with your use of AI generated content. While we do provide some guidelines and suggestions, AI generated content is an area that is in considerable flux right now and these guidelines and recommendations may change as the field evolves.
- Manually review and assess all AI generated content for accuracy, appropriateness, and usefulness before including it in any OER. AI generated content should be reviewed by more than one subject matter expert to ensure the validity of the content. As an OER author, you are ultimately accountable for the content that you share in your OER, therefore you must manually verify the accuracy of the content.
- Closely review any AI generated content for bias, including language or images that reinforce cultural or societal stereotypes around race, ethnicity, colour, ancestry, place of origin, political beliefs, religion, marital status, family status, ability, sex, gender identity and expression, sexual orientation, age, and class and/or socioeconomic status. Consider reviewing and assessing the outputs of AI generated content using the BCcampus EDI statement and ask whether the content aligns with these considerations.
- Do not use generative AI to generate content for an area or subject where you do not have the appropriate level of knowledge or understanding to verify the accuracy of the content.
- Be transparent about your use of generative AI. Just like attributing the reuse of open content, you should include statements within the OER that let others know that you have used generative AI in the creation of the OER. This should include;
- what content was generated
- what tools were used to generate the content, including links to the tool,
- how you used that tool (ie what prompts was the tool given that generated the content)
- the date the content was generated
- what steps were taken to review the content to ensure it was valid and correct.
- As much of the legal consensus around AI generated content suggests AI created content is not copyrightable, you should not apply a Creative Commons license to AI generated content as Creative Commons licenses can only be applied to content that is copyrightable.
- You should avoid generating content that may include content that is protected by a trademark or patent. For example, you should avoid creating an image using an AI image generator that includes a trademarked corporate logo unless you are doing so under the purposes of Fair Use.
- If you use AI to create a summary of another work, you should ensure that you are familiar enough with the original work to determine whether or not the generated summary is an accurate representation of the original work before using the summary.
Is it safe to use ChatGPT image description:
- Does it matter if the output is true?
- No. It is safe to use ChatGPT.
- Yes. Continue to next question.
- Do you have expertise to verify that the output is accurate?
- No. It is unsafe to use ChatGPT.
- Yes. Continue to next question.
- Are you able willing to take full responsibility (legal, moral, etc.) for missed inaccuracies?
- No. It is unsafe to use ChatGPT.
- Yes. It’s possible to use ChatGPT, but be sure to verify each output word and sentence for accuracy and common sense.
Attributions
“When is it safe to use ChatGPT” flowchart by Aleksandr Tiulkanov in ChatGPT and Artificial Intelligence in higher education: a quick start guide [PDF] by UNESCO is licensed under a CC BY 4.0 licence. Flow chart recreated by BCcampus to improve readability.
References
Bommasani, R., Klyman, K., Longpre, S., Kapoor, S., Maslej, N., Xiong, B., Zhang, D., & Liang, P. (2023). The Foundation Model Transparency Index (arXiv:2310.12941). arXiv. https://doi.org/10.48550/arXiv.2310.12941
Gartner. (n.d.). Generative AI: What Is It, Tools, Models, Applications and Use Cases. Gartner. Retrieved July 5, 2023, from https://www.gartner.com/en/topics/generative-ai
Lalonde, C. (2023, March 6). ChatGPT and Open Education. https://bccampus.ca/2023/03/06/chatgpt-and-open-education/
Li, P., Yang, J., Islam, M. A., & Ren, S. (2023). Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models (arXiv:2304.03271). arXiv. https://doi.org/10.48550/arXiv.2304.03271
Sabzalieva, E.,, & Valentini, A. (2023). ChatGPT and Artificial Intelligence: A Quick Start Guide [PDF]. UNESCO. https://www.iesalc.unesco.org/wp-content/uploads/2023/04/ChatGPT-and-Artificial-Intelligence-in-higher-education-Quick-Start-guide_EN_FINAL.pdf
United Nations Educational, Scientific and Cultural Organization. (2023). Guidance for generative AI in education and research (p. 45). https://unesdoc.unesco.org/ark:/48223/pf0000386693
University of Waterloo Copyright Advisory Committee. (2023, August 15). Generative artificial intelligence. https://uwaterloo.ca/copyright-at-waterloo/teaching/generative-artificial-intelligence
Wolfson, S. (2023, February 17). Fair Use: Training Generative AI. Creative Commons. https://creativecommons.org/2023/02/17/fair-use-training-generative-ai/
Wolfson, S. (2023, February 21). This Is Not a Bicycle: Human Creativity and Generative AI. Creative Commons. https://creativecommons.org/2023/02/21/this-is-not-a-bicycle-human-creativity-and-generative-ai/