The Ethics of Generative AI in Learning and Development

Using generative AI tools in the field of learning and development can sometimes feel like we’re back in the wild American West. With easily accessible models like Claude and ChatGPT, an instructional design assistant is always at our beck and call. However, with little to no formal legislation and even fuzzier usage guidelines, how should we navigate these tricky waters of ethics and fair use?

In the mid- to late- 19th century, miners discovered large deposits of precious metals in the mountains of the western United States. The ensuing “rush” to capture wealth and riches brought a sea of people from all corners of the world. Along with this migration came chaos, lawlessness, and confusion. Cities like Nevada City, Tombstone in Arizona, and even Park City here in Utah sprang up almost overnight. As they grew, their residents realized the need to come together and create laws, principles, and governments to keep their communities safe and prosperous.

Taking inspiration from our resilient forbearers, our team has decided to govern ourselves by creating a set of ethical rules we choose to follow for using generative AI as an instructional design and thought partner. I don't suggest these guidelines are right for everyone in our industry, but they work for us currently and help us mitigate largely unintentional risks—both those we create and those we’re subject to. I've chosen to share them here because I would love feedback, and I want to open a dialogue so we can all better understand what is working even more broadly for us as a community.

Guideline #1 – Without explicit permission or license, we never use AI to create instructional designs that convey a thought leader’s core models, research, or insights.

This guideline is our true north. Intellectual property ownership and copyright protection can sometimes be gray areas in our industry. Even though it’s widely understood that intellectual property created before 1923 is reliably in the public domain, we occasionally overlook this fact due to the widespread dissemination of ideas and content on the internet and in other publicly accessible repositories of information. To ensure we respect the rights of our many thought leader partners and friends, we employ AI to support primarily only those learning experiences and simulations that are ethically implemented and for which their creators have been fairly compensated.

Guideline #2 – We never upload copyrighted materials into a generative AI platform unless we have the copyright holder’s permission and are operating in a secure environment as AI users.

Unless you dig a few levels deep, it’s not well publicized that any content uploaded into an AI model may be retained within that model and used for its future training and interactions. In many large organizations that are AI-savvy, there exists an internal model protected by firewalls, encryption, and other security measures that effectively isolates uploaded data from the broader AI landscape. For smaller practitioners like us, most models—such as ChatGPT, which we use—offer a paid team subscription service that provides us with a private, secure environment. The only thing more disastrous for us than unintentionally sharing our intellectual property with the world would be to inadvertently expose our clients’ confidential data in the same way.

Guideline #3 – All AI-created content is considered a first draft that requires human review, revision, and voice.

On any given day, AI can either be the most capable thought partner we’ve ever worked with or a frustrating timewaster. We are humans using AI to help create content for other humans. Only we can know the right tone and inflection with which someone might read a sentence or contextualize a scenario. Our voice embodies our human spirit, and it is incredibly challenging for any current AI tool to emulate that voice accurately. Additionally, as we consider our core values as both humans and as an organization, we believe it is important to demonstrate integrity and accountability in reviewing outputs generated by AI for both inaccuracies and truth.

Guideline #4 – We use AI for creativity and insights.

We approach every interaction with an AI model as an opportunity to improve, not simply to increase efficiency. We believe this intent helps us unlock the power and promise of AI in our field. While efficiency is valuable and something we’ve gained through increased AI usage as a team, focusing on this alone detracts from our three-part value proposition of being creative, insightful, and accountable to our clients.

This post isn’t intended to be a manifesto; it’s the beginning of a conversation we hope will spark curiosity among L&D professionals about how we can come together to discuss governing ourselves as a broader community. Likewise, these guidelines aren’t immutable for us. We consider them dynamic, and as we learn through usage and from others, we will adjust or add clarity to make sure we’re evolving alongside this special technology.

If any of you are willing to join this conversation and share your thoughts and insights, I welcome them in the spirit of learning and exploring this new frontier together. At the end of the day, I think we all know, “There’s gold in them thar hills!”