In the Age of AI, engineers still need to learn how to write. They need to learn to use writing to develop ideas, record findings, document decisions, and persuade colleagues, both as students in our curricula and also as professional engineers (Berdanier and Alley, 2023). Writing produced by GenAI cannot replace the writing process and its role in how we develop our thinking and organize information into meaning, though GenAI platforms offer powerful potentials to support the writing process. GenAI also poses a number of important limitations, including making up facts, and getting literature references incorrect (Eaton, Mindak, and Morrison, 2021). 

As faculty in the College of Engineering, one of our responsibilities to students is to guide them to ethical and developmentally appropriate uses of writing with GenAI. The following guide offers an overview of writing in the age of AI as well as practical strategies for limiting students' uses of GenAI in writing assignments as well as incorporating GenAI into writing assignments.    

Writing in the Age of AI

Many of our colleagues have expressed concerns about GenAI, especially regarding its potential use for plagiarism and other shortcuts to learning. These attitudes are not new: at least since Plato’s Phaedrus, each successive technological development in the history of writing and literacy has been met with concerns about how and whether the emerging technology of the day is going to ruin human thought, human development, and perhaps society in general. Those concerns are often overstated, and today’s moment is likely no different. However, each new major writing technology (paper, printing press, ballpoint pen, personal computing, the world wide web) has affected how we think and make meaning with language, and how we connect with each other in larger economic and political groups. GenAI will be no different, but what those disruptions are or will be has yet to be fully determined. 

So what is writing in the Age of AI? The answer to this question is still unfolding, but mostly, writing is still what it has always been. 

  • Writing is a social and rhetorical act, as we write to fulfill our communicative purposes using genres in order to reach target audiences in specific situations (Lunsford, 2015). 

  • Writing is a mode of thinking and learning that allows us to discover, develop, and articulate ideas (Estrem, 2015; Dryer, 2015). 

  • Writing is a process (Yancey, 2015) with a beginning (planning), middle (composing and revising), and end (editing and publishing). 

  • Writing happens through specific genres (Bazerman, 2015; Hart-Davison, 2015) that allow us to easily communicate to each other in communities, like lab reports written for a client. 

  • Writing is a technology that mediates language with and through devices, instruments, and materials (Brooke and Grabill, 2015), including computers and GenAI. 

  • Writing is composed by writers (who interact with various tools, including GenAI to develop language) as they enact specific social identities through language (Estrem, 2015; Gee, 2015), like being an engineer, thus taking authorship or responsibility (Cooper, 2011) for the message expressed in the text. 

  • Writing allows people to complete socially meaningful tasks or actions (Russell, 2015), like redesign a prototype with a large team of engineers. 

These threshold concepts of writing are well established in writing and language disciplinary research, and GenAI is unlikely to change the fact that these principles operate in any writing event, although emerging and evolving technologies will likely impact how our students complete these tasks. 

We’ve been writing with robots for a long time now (spellcheck, grammarcheck, Matlab) (Hart-Davison, 2020), and our current moment represents the next phase of the human-technology interface (Duin and Pedersen, 2021). However, writing is both a cognitive and social practice, which means that the writing practices that we regularly use impact how we think and how we distribute our cognition through the tools and materials involved in writing (Bransford, Brown, and Cocking, 2000; Wolf and Barzillai, 2009). 

Strategies for Limiting Students’ Uses of GenAI

There are several reasons faculty may want to limit students’ uses of GenAI in the classroom. From a developmental standpoint, Tate et al. (2023) observe that “Teachers will have to balance the teaching of effective writing with AI and writing without AI to ensure that students build up the necessary ‘muscle tone’ to write and do not move too quickly to AI-generated text. Writing is hard; thinking is hard. But there is a time and a place for practicing hard things in order to become more proficient at them” (10, emphasis added). Our responsibility as educators is to figure out when is the best time to incorporate these powerful tools into our students’ workflows in ways that don’t stymie the hard cognitive work they will have to do with writing. It can be easy for us, experts in our own fields with years or decades of experience writing for disciplinary audiences, to see GenAI as an important resource for saving some cognitive energy. But for our students, and especially the least experienced of our students, those cognitive challenges are important experiences that will enable them to use GenAI effectively and strategically. 

Students will not always have access to computers to mediate their ideas. While GenAI might help students develop and articulate their ideas, without a deeper understanding of  concepts, students (both in class and as practicing engineers) may not have the cognitive readiness to develop, discuss, and represent their ideas in other settings, such as oral communication that requires thinking on their feet. The writing process helps people select, develop, and organize their thoughts, thereby structuring the learning process required to holistically understand an issue. 

  • Focus on writing processes: GenAI can produce well written text that is generally free of grammatical errors and follows language conventions. Consider de-emphasizing this part of your writing assignments by creating low stakes assignments (i.e., credit/no-credit submissions) that support larger projects. Focus on Process over Product by asking students to submit lab notes that support their memo writing, or by drafting the methods section in class at the end of their lab. By requiring and giving credit for these lower-order tasks, instructors will disincentivize use of GenAI to complete assignments. 

  • Create assignments that emphasize higher-order thinking, reasoning, and metacognition. GenAI is really good at generating text in response to a specific, limited question. Ask students to think at a higher level, to explain the reasoning they use to solve a problem. You may choose to formulate these as reflective writing assignments that ask students to explain key concepts related to the required subject matter knowledge and how they used that knowledge in application to a specific problem. Emphasizing these types of questions can further support students’ ability to transfer their knowledge to other situations (Robertson and Taczak, 2017).

  • Design assignments that require rigorous engagement with sources, especially journal articles: Most GenAI platforms, including ChatGPT, are not yet equipped to work from sources in a rigorous way and are not reliable in their use of sources and citations. This weakness often manifests as fabricated or erroneous sources, which creates a significant issue for peer-reviewed articles. Consider structuring assignments around rigorous use of sources and appropriate citations practices to disincentivize students from relying too much on chatbots for their responses. If students use chatbots to generate preliminary bibliographies, require them to check each citation for accuracy (and to make sure the source exists). You may also want to require students to incorporate sources very early in drafts in order to encourage them to build much or all of their writing around the reading work they did in the research. The goal here is to disincentivize students from overly relying on chatbots to generate their responses, while also engaging students in careful reading, analysis, and synthesis from sources.

  • Lean on class-specific content and context: Create assignments that rely on class-specific discussion and content. Because ChatGPT is trained on large data sets of language, it isn’t as effective at generating responses to class-specific discussions, localized contexts, or recent events. When possible, tie writing assignments to design work or lab work completed by students in the course. This material will be difficult for GenAI to easily reproduce or reference without substantial effort from the student writer.  

Strategies for Supporting Students’ Uses of GenAI 

Before we look at specific strategies, we should keep two points in mind. First, select strategies that support your larger goals for having students use writing to think (and demonstrate learning in the classroom). Note: many of our students have been talking to algorithms (e.g., Alexa and Siri) for much of their life and are becoming increasingly adept at using GenAI to assist them in their writing assignments (Svrluga and Natanson, 2023; Terry, 2023), so articulate for students how the use of GenAI supports your course’s larger learning outcomes. 

Second, GenAI platforms are just word predicting machines–so while the language generated may seem to be coherent, there is only an illusion of meaning (Bender, et al., 2021). Meaning in language comes out of the push and pull of writers’ intentions and readers’ interpretations of a text. While the language generated by GenAI is prompted by human writers, the meaning of the text depends on a human author asserting responsibility for what has been said through language and inserting that text into social situations where others’ actions require the use of the text itself (e.g., reading a lab report in order to develop a design change). GenAI can help writers develop messages, but GenAI cannot replace writers and the communities of action in which they are embedded. 

  • Writing Process Strategies - knowledge about how to develop a text

    • Assign students tasks that help them invent ideas for a written project. For example, if students have to do a research project that draws on secondary resources, you might have students develop an outline before they draft. Alternatively, you can have students write a draft, then ask GenAI to draft an outline based on the rough draft composed by the student. This outline could then help the student develop the text to the next level. 

    • Revise prose for specific audiences. You can have students take a technical prose and revise it for a general (public) audience (or vice versa) using GenAI. Similarly, students can insert a section of their writing and ask GenAI to help improve the thesis statement or topic sentences in the section. 

    • Edit prose before submitting a final draft. Students can insert text to GenAI and ask the machine to edit the prose for any grammar and/or spelling mistakes. Note that much of the text’s formatting will be lost (e.g., bold, underlining, headers, etc.) in the conversion into and out of the chatbot. Note also that the chatbot will eliminate features of minoritized English dialects, replacing students’ home languages with White Mainstream English and thereby revising their expression of social identity. 

  • Subject Matter Strategies - knowledge rooted in engineering disciplines

    • If students haven’t taken a prerequisite course, they may be able to use GenAI to start to develop required background knowledge, or to check their comprehension of certain concepts. Of course, this is an area where the factual reliability of the algorithm may be a problem. 

    • Develop code for MatLab and other coding needs to support projects in the lab or design processes. There is a steep learning curve for developing code for MatLab and for developing code that can help students solve various data collection or analysis requirements. GenAI can often draft the code needed for students or help debug this code, depending on the situation. Note that some basic knowledge of coding will be required for these processes to work, though GenAI can generally provide the information needed to solve most problems. 

  •  Genre Strategies - knowledge about text types (e.g., lab report, grant proposal)

    • Compose broad pieces like a summary or abstract. Once the student has written a full draft, they can input that draft into GenAI and ask the machine to return a brief summary or abstract, or perhaps a list of key terms associated with the text.

    • Input the list of instruments used and a general outline of the research methods used in lab and ask GenAI to draft a methods section. Students should treat this output as a first draft. 

    • Transition one genre into another by inputting the draft of a lab report into GenAI and asking it to rewrite the content into a public letter (or some other technical or non-technical genre) to a city council that makes the same argument as the lab report. 

  • Rhetorical Strategies - knowledge about connecting purpose with audience

    • Tell GenAI your purpose for writing a given text and ask the algorithm to assess whether a text sample clearly articulates that purpose to the audience. 

    • Ask GenAI to evaluate whether you explain why a piece of evidence supports the stated claim in a limited text sample. 

    • Take a text written for a technical audience and ask GenAI to recommend revisions so that it can be understood by a nontechnical audience.

Next Steps 

This document has provided some general guidelines for conceptualizing the relationship of GenAI with writing and how you might approach supporting or limiting its use in your classroom. Here are some possible next steps:

  1. Spend some time exploring GenAI. You might want to consider using UM’s GenAI, UM-GPT (https://umgpt.umich.edu/) and just asking the algorithm to respond to prompts. Try putting in your writing assignments or scaffolding questions students might ask to see what kind of results you get. Explore and play around with the technology, and then talk with your colleagues about what they have experienced. Consider how you might use GenAI in your own workflows, and think about how those opportunities might inform your teaching practices. 

  2. Talk with your students about their experiences with GenAI. Strike up conversations before class begins or in the hallway as you wait for your room to become available. What do they think? How have they been using the technology? What are the strengths, limitations, and/or frustrations for them? Share those insights with your teaching partners and departmental colleagues.

  3. Read about how others at institutions across the country and across the disciplines are thinking about this technology. Look at conversations related to GenAI and writing as well as how it is being applied to support student learning of disciplinary knowledge in engineering. Check out “AI Text Generators,” edited by Anna Mills, an exhaustive list of popular press and academic press resources on GenAI in all facets of higher education. Share the resources you find helpful with students and colleagues. After all, this is an issue that all of us need to grapple with. 

References

Bazerman, Charles, (2015). “Writing speaks to situations through recognizable forms.” Naming what we know: Threshold concepts of writing studies, Ed. Linda Adler-Kassner and Elizabeth Wardle. Utah State University Press, 35-37.

Bender, Emily M., et al., (2021) “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜,” in Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, FAccT '21 (New York: Association for Computing Machinery, 2021), 610–23.

Berdanier, Catherine G. P. , and Alley, Michael. (2023). “We still need to teach engineers to write in the era of ChatGPT.” Journal of engineering education, 583-86.

Bransford, J. D., Brown, A. L., and Cocking, R. R. (2000). How people learn (Vol. 11). Washington, DC: National academy press.

Brooke, Collin, and Grabill, Jeffrey T. (2015). “Writing is a technology through which writers create and recreate meaning.” Naming what we know: Threshold concepts of writing studies, Ed. Linda Adler-Kassner and Elizabeth Wardle. Utah State University Press, 32-33.

Cooper, Marilyn.  (2011). “Rhetorical agency as emergent and enacted.” College Composition and Communication, 62 (3), 420-49.

Dryer, Dylan. (2015). “Writing is (also always) a cognitive activity.” Naming what we know: Threshold concepts of writing studies, Ed. Linda Adler-Kassner and Elizabeth Wardle. Utah State University Press, 71-74.

Duin, Ann Hill, and Pedersen, Isabel. (2021). Writing Futures: Collaborative, Algorithmic, Autonomous. Studies in Computational Intelligence, v. 969. Springer. https://doi.org/10.1007/978-3-030-70928-0

Eaton, Sarah Eliane, Mindzak, Michael and Morrison, Ryan. (2021). “Artificial intelligence, algorithmic writing & educational ethics.” [Paper presentation]. Canadian society for the study of education Société canadienne pour l’étude de l’éducation, Edmonton, AB, Canada. http://hdl.handle.net/1880/113569 conference paper.

Estrem, Heidi. (2015). “Writing is a knowledge-making activity.” Naming what we know: Threshold concepts of writing studies, Ed. Linda Adler-Kassner and Elizabeth Wardle. Utah State University Press, 19.

Gee, James Paul. (2015). Social Linguistics and Literacies: Ideology in Discourses, 4th Ed. Routledge.

Hart-Davidson, William. (2018). “Writing with robots and other curiosities of the age of machine rhetorics.” In The Routledge handbook of digital writing and rhetoric, Ed. Jonathan Alexander and Jacqueline Rhodes. 248-255.

Robertson, Diane, and Taczak, Kara. (2017). “Teaching for Transfer” in Understanding Writing Transfer: Implications for Transformative Student Learning in Higher Education Ed  Jessie L. Moore and Randall Bass. Stylus Publishing, Sterling, VA. 93-102.

Roozen, Kevin. (2015). “Writing is a social and rhetorical activity.” Naming what we know: Threshold concepts of writing studies, Ed. Linda Adler-Kassner and Elizabeth Wardle. Utah State University Press, 17-18.

Russell, David R. (2015). “Writing mediates activity.” Naming what we know: Threshold concepts of writing studies, Ed. Linda Adler-Kassner and Elizabeth Wardle. Utah State University Press, 26.

Svrluga, Susan, and Natanson, Hannah. (June 1, 2023). “All the unexpected ways ChatGPT is infiltrating students’ lives.” The Washington Post. 

Tate, Tamara, Doroudi, Shayan,Ritchie,  Daniel, Ying Xu, and Warschauer, Mark. (2023). “Educational research and AI-generated writing: Confronting the coming tsunami.” Preprint.

Terry, Owen Kichizo. (2023). “I’m a Student. You have no idea how much we’re using ChatGPT.” The Chronicle of Higher Education, May 12. 

Wolf, Maryanne, and Barzillai, Mirit. “The Importance of Deep Reading: What Will It Take for the Next Generation to Read Thoughtfully – Both in Print and Online.” Educational Leadership 66.6 (2009): 32-37.

Yancey, Kathleen Blake. (2015). “Learning to write effectively requires different kinds of practice, time, and effort.” Naming what we know: Threshold concepts of writing studies, Ed.Linda Adler-Kassner and Elizabeth Wardle. Utah State University Press, 64-66.