The University of Wisconsin-Madison Division of Extension is committed to using generative AI responsibly in a way that aligns with our values and upholds our commitment to providing the public with research-based information. This memo is meant to support staff as they incorporate AI into their work. These principles follow UW-Madison policy and are informed by best practices from peer organizations like Wisconsin Public Media (WPM).  

Guiding Principles 

• Embrace Innovation: We see the potential of generative AI to enhance our work and will approach this technology with a spirit of curiosity and a commitment to responsible use. 

• Human Oversight: We maintain that AI is a tool that functions best when partnered with human oversight. We are responsible for the content we create, even when using AI.   

• Research-Based Information: Our priority is to provide information based on solid research. AI does not replace our responsibility to use reliable sources and methods. 

Best Practices 

• Follow established rules. We will follow UW-Madison's rules for keeping data secure when using generative AI. 

• Don’t be afraid to talk about it. We believe in being transparent about our use of AI, both to champion and model AI use and adoption and to foster trust in our content. Consider using citations or attributions when sharing wholly AI-generated content with the public (such as imagery).  

• Think about bias. We are aware that AI systems can sometimes have biases. Using human oversight to review content for bias can mitigate the effects of bias. For example, ask a colleague to review your output (and even your prompts). How we ask an AI for help can impact what we get back. Our colleagues at the Office of Access, Compliance and Inclusion can be a resource when we have questions. 

• Stay curious! AI is rapidly changing. We believe in a growth mindset around AI and encourage ongoing learning. What will AI do next? 

• But don't overdo it. It’s easy to push AI to its limits. Consider asking yourself, “how do I know this information is accurate?” If you can’t answer that question, maybe AI is not the right tool for the job.