It would only be appropriate that a toolkit that touts the benefits of AI and recommends that “its use must be disclosed and explained” would have a section dedicated to explaining how AI was used in its development. In this section, we explain how AI was used as a writing and research assistant, and rather than adhere to evolving approaches to citing and referring to generative AI text (see APA, Chicago, and MLA), we’ve chosen to share our approach by providing example prompts as guidance for others.
Multiple generative AI tools were used to help refine this toolkit’s text, including Bard, Bing Chat, ChatGPT 4.0, and Claude. AI tools were primarily used to transform initial concepts into draft sentences for further refinement. For example, tools were frequently prompted to “Clean this up:” accompanied by a rough attempt at a sentence. Assuming the output was satisfactory, adjustments were made to fit the context of the surrounding text. These revised sentences were frequently re-inputted into the AI tool to be shortened or to remove technical jargon. Every output was thoroughly evaluated before it was included in the toolkit.
Below are examples of tasks, prompts, and actions based on text created by generative AI tools.
Why use more than one generative AI tool?
While cost, familiarity, performance, and how recent the training data is were all factors in choosing tools, the same prompt was often inputted into two or more generative AI tools to simply get more variety in the outputs. Portions of output from each tool were often combined to create a final text.
Not all of the AI tools used to write the Toolkit were generative AI tools. Grammarly, an intelligent writing assistant that uses advanced natural language processing, was used to detect and suggest spelling and grammar edits. These were each evaluated and then accepted or rejected.
We want to acknowledge the importance of human creativity and collaboration in creating this toolkit. Our toolkit could not have been created without the other humans who envisioned, discussed, drafted, reviewed, and revised the resources. Not all of the co-authors decided to use AI tools. Effective collaboration across organizations, which included sharing ideas, examples, expertise, and learning from one another, led the development of this toolkit while generating words to help express the ideas was the main contribution of AI.
AI Guidance for Schools Toolkit © 2023 by Code.org, CoSN, Digital Promise, European EdTech Alliance, and PACE is licensed under CC BY-NC-SA 4.0