A (Working) Framework for AI Literacy
Licensed under CC BY NC - heplerconsulting.com
NOTE: This is by no means a perfect framework, but it is a list of the central skills, concepts, and ideas that I feel are necessary for one to be considered “AI-literate” in the truest sense. The CC BY NC license means that you can copy it in its entirety and adapt it any way you want, as long as you link to this original site and do not use your derivation for commercial use. For more information on CC licenses, visit Creative Commons and opencontent.org
NOTE #2: I put this framework here to help foster a critical discussion of the central elements of literate and ethical use of generative AI tools, whether in education, professional, or private environments. I will be discussing it at the AECT 2024 International Convention and other locations. Feel free to reach out to me on LinkedIn or at heplerconsulting.com.
Why Now?
I have been asked by several people to release my proposed AI Literacy Framework. I was going to wait until next week to do this, but now that OpenAI has released GPT4o to the public, and thus all Custom GPTs, it is important that we begin this conversation sooner than I previously thought.
Why Do We Need To Talk About AI Literacy?
Many instructors and professionals are discussing best practices for using generative AI tools. They are also debating on which tools are the most and least ethical. Furthermore, some users of generative AI are struggling to understand which tools are most appropriate for which tasks. Novice users are not knowledgeable concerning how much they should edit or change AI products. Issues regarding copyright and information literacy are not fully appreciated by all users. Competition between commercial and open AI tools further clouds the waters for users, educators, and professionals alike.
There must be a framework for AI Literacy that is device-, tool-, and media-agnostic. This framework must ensure that users are knowledgeable about the nuances of AI types, organizations, tools, products, best practices, and ethical issues. AI Literacy should be clearly defined so that no student or novice user participates in AI-supported creation without considering all of its implications.
The Framework
Presented below is my basic framework for thorough AI Literacy:
Attributes
AI Literacy is demonstrated through three main attributes:
1. possessing education-, training-, and experience-based knowledge regarding AI best practices and ethical and social issues,
2. actively applying this knowledge in all interactions with an AI tool or an AI-generated product, and
3. proactively teaching others about these issues, skills, abilities, and paradigms.
Subjects and Abilities
The three attributes above should be demonstrated in the following subjects and abilities (to an appropriate level of depth for each grade, training, or course):
1. machinations and functions of generative artificial intelligence
2. differences between genAI tools for various media
3. proper use of multimodal genAI tools
4. best practices for creating effective, thorough, ethical, and conscientious prompts for genAI tools
5. critical thinking and information literacy in critiquing and verifying genAI products and outputs
6. relationship between AI tools and users, including proper attribution of genAI tools
7. using multiple AI tools in a single workflow or to fulfill a task for which there is no existing workflow
8. ethical, social, societal, and moral issues and questions regarding AI training, function, use, development, attribution, and commercialization
9. abilities and limitations of generative AI tools (being able to decide which tool to use and how)
10. AI feasibility (a term I have come up with to describe the mental process of determining if an AI tool really needs to be used, or if a human can do it faster and more efficiently)
11. utility of various prompt types for different tasks (should I use the GovTech Singapore COSTAR Framework, or will a simple keyword analysis prompt suffice?)
12. personal, group, and institutional responsibility for AI use, outputs, and acknowledgment (this is a growing idea, and one that has not fully been explored)
13. implications, effects, and hazards of genAI tools related to data privacy and confidentiality (see Danielle Keats' Privacy Harms article for her thorough and insightful framework)
14. (for educators) ethical, social, and educational implications of using, trusting, and relying on AI detectors
14. (for students) ethical, social, and educational implications of overreliance on AI tools
An AI Literacy Course
I had the great honor to work with the AI Committee at the College of Southern Idaho to create an AI Literacy Course that we have released under a CC BY NC SA license on Canvas Commons. That course is derived from an original course created by Rush University. Feel free to use this course however you would like, again under the regulations of the license. This literacy course does not follow my framework, partly because I was still creating it when the course was developed and partly because I did not want to force my detailed framework on college students.
Implement the Framework
OpenAI has just released Custom GPTs to ALL users! This means that all of my tools are available. If you would like to test them and provide feedback, just feel free to go to "Explore GPTs" and search my name, "reed hepler." They should all come up, from my Gilbert and Sullivan-lyric quoter, "SavoyGPT," to my OpenOpus Chat Tool that queries classical music metadata.
Another environment that holds links to especially helpful tools (especially for librarians, educators, and researchers) is libraryrobot.org. Steve Hargadon and I released this completely today, which is fortunate because of the GPT4o public release. These tools include a Book Finder, a Book Summarizer, and a Search Query Optimizer (which I have linked to in previous posts). Let me know how these work for you!
GPT4o is a multimodal tool, which means that it can output multiple outputs. It has a text transformer and audio and image diffusers combined into a network that acts as one tool. However, like I said above, this framework is tool- and format-agnostic. The concepts should be applicable no matter what type of AI tool you use.
Reed, on the subject of AI Feasibility, a good term btw, The keynote speaker at the University of Missouri-Columbia's recent Celebration of Teaching Conference, Regan Gurung, suggested that we need to evaluate when to use or not use an AI based on what he called the FEAL framework:
Faster - can AI help us do the job faster, or is our skillset such that it would hinder us?
Ethical - can we use AI responsibly for a given task?
Accurate - how accurate is the AI in performing this task?
Learning - does using AI for a task help or hinder our learning?
I find that a useful start.
Of your other points, would you clarify what you mean by "machinations" in "machinations and functions of generative artificial intelligence" - that is unclear to me, particularly since machinations has a sinister connotation to me.
My biggest criticism of most of the AI literacy frameworks is that there is a lack of any larger contexts. I understand that they have to be focused, but I also believe they have to be embedded in an understanding of the history/sociology of technology and also in a broader understanding of intelligence, consciousness, and sentience across the natural world - a field of study that is constantly expanding and should affect both how we understand and use AI.