Department of English

Generative AI does not produce writing. Generative AI systems that are Large Language Models (LLMs) produce tokens or symbols by predicting likely language patterns learned from large datasets. We comprehend these tokens or symbols as text. LLM outputs tend to be predictable and homogenized and may reproduce or amplify existing social biases. LLMs do not think, feel, interpret, or take responsibility for the text they generate. Because they do not verify accuracy, they can produce false or fabricated information. AI-generated text should be treated as a tool—not an example of writing authored by a person or a source of knowledge. 

We acknowledge that generative AI has the potential to be a transformative technology in many ways, but we have concerns about current labor conditions, environmental costs, training materials, and invasive and toxic marketing.   

• Environmental costs: building and maintaining data centers to host generative AI systems require massive amounts of energy (to power) and water (to cool), contributing to carbon emissions production and depletion of the clean water supply. 
• Labor conditions: the human workers who train generative AI systems on data sets are poorly compensated for their work, and they face job insecurity, lack of training, language barriers, and isolation that keeps them from knowing what the consequences of their work will be.   
• Training materials: many of the training materials used to create data sets are accessed in unethical and illegal ways that do not compensate the creators of those materials.   
• Marketing generative AI to students as educational technology: students are repeatedly invited to use generative AI tools with the promise that they will improve their success in their classes. These marketing materials interrupt students during their writing process, offering tempting access to assistance, but we caution students to critically analyze this messaging and these services.

We are concerned that they: 

• undermine students' confidence in their own abilities, preying on students’ insecurities about their writing and promising them that a product like Grammarly will make them successful.   
• suggest the fundamentals of writing are annoying tasks to be elided as quickly as possible, which directly counters the value that our department places in the difficulties, struggles, hard-won realizations, and unpredictable pathways to success that are inherent in the writing process. 
• present themselves as a suitable alternative to a learning community of peers, instructors, and researchers. Anything social may already intimidate students, and the gen-AI messaging tells them that they don't need to bother anyway, because Grammarly or Chat GPT is there for them. 

Faculty members may choose whether to permit the use of generative AI tools in their courses or assignments. When AI use is allowed, students must disclose it fully. Instructors may require one or more of the following:

• A brief narrative explaining how AI was used during the writing process.   
• A transcript of all prompts and all AI-generated responses.   
• A reflection on how AI helped or hindered the development of the work. 

If an instructor suspects unauthorized or undisclosed AI use, the student may be asked to:   

• Discuss their writing process and the ideas within the submitted work.   
• Provide drafts, notes, and/or other evidence showing the development of the assignment. 

If an instructor uses AI to evaluate student writing, that activity must be disclosed to students, including a description of what AI was allowed to do.