Composing with AI

Pedagogical Implications

In this opportune moment, we would like to take a step back to reflect on how this exploratory usability study might inform the teaching of writing. Incidentally, each of us is also a writing teacher and we are keenly interested in pedagogical implications. Indeed, we agree with the many writing scholars who suggested that ChatGPT provides an opportunity to “re-see” our writing pedagogy. In that spirit, and as writing teachers ourselves, we note some initial pedagogical implications for ChatGPT in writing classrooms. These implications include expanding writing pedagogy to integrate generative AI in writing processes, developing critical AI literacy, and addressing ways that generative AI impacts textual integrity and plagiarism. We also advocate a critical pedagogical perspective for implementing Generative AI technologies in writing classrooms. In the sections below, we outline each of these areas more fully.

Connecting Generative AI and Writing Processes

As we noted, findings of this study suggested that students appreciated the capabilities of ChatGPT for process-related activities including generating ideas, brainstorming, and outlining. Yet, students also articulated a number of questions about ChatGPT texts such as doubts about information credibility, lack of depth, and ethical concerns about using ChatGPT. In response, we advocate creating opportunities for students to explore ChatGPT for process activities, but we also suggest centering a critical perspective in doing so. What might this look like? Pedagogical activities related to writing process could involve facilitating all-class activities or exercises in which students are invited to use ChatGPT at various stages of their writing processes. The whole class approach is helpful for ensuring that all students have the opportunity to interact and critically think about ChatGPT. Students could be asked to work in groups to explore ChatGPT, document its suggestions, and reflect on what is most useful and what is not useful about ChatGPT contributions. Potential prompts to ChatGPT might include brainstorming ideas on a specific topic or generating outlines. In these exercises, reflection is a key part, as it provides the opportunity for students to critically analyze ChatGPT as they have experienced it. Student groups might point out weaknesses and strengths of ChatGPT suggestions in all-class discussion. Table 4 includes specific suggestions for various process-based class activities that carry this critical perspective.

In addition, and as Graham suggested, ChatGPT strangely provides an opportunity for us to expand our understanding of process-based writing through a “post-process” lens that recognizes the unique attributes and functions of ChatGPT. According to Graham, a post-process pedagogy involving generative AI would view writing as an interaction with LLMs corpus indirectly, and could center on the practices of fact-checking, curating, prompting, and revising. These practices could be built into classroom activities and assignments, such as by encouraging students to “fact-check” responses produced by ChatGPT, or examine the ways different prompts generate different responses, all with a critical eye. We will return to this perspective in the next section. Ultimately, however, ChatGPT teaches us that critical AI pedagogy is to be centered–and not merely affirmed—in writing processes.

Practicing Critical AI Literacy

Again, we find a critical perspective both helpful and necessary in thinking about pedagogical implications of using ChatGPT. Critical pedagogy centers on the notion of critical consciousness, and here, we consider what a “critical AI literacy” might look like. Such a stance would be open to an explicit awareness of generative AI technologies and would simultaneously embrace a critical stance and perspective.

Two concepts come to mind that may help establish a “critical AI literacy.” One is to consider technologies and ethics simultaneously. In Augmentation Technologies and Artificial Intelligence in Technical Communication: Designing Ethical Futures, Duin and Pedersen (2023) present an “ethical futures framework” that would “reframe professional practice and pedagogy to promote literacies surrounding the ethical design, adoption, and adaptation of augmentation technologies” (“Stronger Relationships, Stronger Programs: Asserting expertise for a generative AI Landscape”, 2023). This framework involves (1) knowledge of augmentation technologies (2) plans for building literacy capacity and (3) strategies for integrating ethical perspectives of adoption and practice. A second example concept involves a “post-digital critical pedagogy,” a pedagogical concept aiming to address the gap between designers and users. In “Postdigital Critical Pedagogy,” Petar Jandric and Sarah Haye (2021) described how this gap can exist: “...tech people did not care about critical pedagogy, and critical pedagogy did not care about tech” (p. 324). They advocated for designers and users to work together to better understand the internal workings of technology and to ask critical questions of it.

Both of these examples help us imagine “critical AI literacy.” Adopting an ethical framework in relation to ChatGPT might mean working to understand the technology and its capabilities while openly acknowledging the ways its use requires scrutiny and verification. In our study, students asked a number of critical questions about ChatGPT, such as “How does it pull information from the internet?”, “How much more advanced will this be in a couple of years?”, “Does it generate a different response for everyone?”, “What kind of database is it pulling from?”. These questions showed that student participants are curious about the authenticity, credibility, transparency, and trustworthiness of AI. We see such questions as productive starting points for practicing “critical AI literacy,” and we encourage writing instructors to create spaces and activities in which students can ask these questions.

One exercise might be to ask students to apply an ethical framework to ChatGPT by discussing how and in what situations various ChatGPT texts would be acceptable. Another exercise might include fact-checking and citation checking exercises which require students to use ChatGPT but also identify and check assertions and citations used in ChatGPT texts. This activity encourages critical reflection about ChatGPT assertions and citations that are seemingly confident; in checking assertions, students can directly question the accuracy of information included in ChatGPT texts. Another suggested exercise is to create requirements or rubrics that articulate criteria for ChatGPT texts. By creating requirements or rubrics that more explicitly describe quality concerns that they are looking for, students using this technology may notice a) that ChatGPT likely cannot produce the depth they are looking for and therefore would not be worth copying word for word and b) prime them to engage certain critical thinking skills that they may not naturally bring to texts produces by AI. Because ChatGPT texts are initially impressive and appear relatively polished on the surface, encouraging a dive beneath the surface could be pedagogically beneficial.

Developing Textual Integrity and Authenticity

Textual integrity and authenticity directly address concerns of plagiarism and ChatGPT attribution, as well as policies governing ethics of ChatGPT use. Regarding textual integrity, Dobrin (2023) focused on authenticity as a guiding concept and encouraged individual students to determine the ways in which their texts reflect their unique and authentic work. We are reminded as well of Duin and Pedersen’s (2021) explanation of the ways that autonomous AI technologies can work in tandem with human voices. Recall that in our study, comments coded in the filter of “self and experience” reflected student thoughts about how they would imagine integrating their unique voice into texts produced by ChatGPT, by expanding ideas, adding citations, changing organization, or other revisions. Students also clearly saw ChatGPT as a potentially helpful tool for more low-order concerns such as help with grammar, citations, or other mechanical writing issues. While other AI tools like Grammarly, Essay Help, and other language booster tools, are already used by students, ChatGPT seems to require a stronger critical consciousness. We acknowledge that ChatGPT also reinforces Standard English and in so doing diminishes the linguistic diversity that students would bring to the classroom. Pedagogical exercises might help students think through what that intermingling looks like and what would be considered acceptable. It might be useful for students and instructors to think critically about that threshold through practice examples of ChatGPT integration. In this spirit, we suggest pedagogical approaches that encourage students to focus on authenticity, including asking questions about how best to integrate ChatGPT information. Should ChatGPT contributions be cited in student writing? What does it mean to heavily edit ChatGPT texts? To what extent is ChatGPT a collaborative writer? Exercises in Table 4 provide suggestions for exercises that focus on these directions.

In addition, instructors may want to call attention to attributive uses of ChatGPT that may be more blatantly unethical, such as taking an entire ChatGPT text and submitting it for homework or a paper. Recall that in our study, student participants shared that they were not likely to use ChatGPT unaltered. Yet there may be situations in which students would wonder if such uses of ChatGPT are acceptable. Following Dobrin, we suggest that instructors clarify, to the best of their ability, how they would interpret such uses of ChatGPT. Instructors may also review their institution’s suggestions on ChatGPT usage. At our institution, our Faculty Senate approved of faculty taking one of three approaches with ChatGPT: (1) encourage usage without reservation (2) encourage limited usage and with reflection (3) prohibit usage of ChatGPT. Whichever scenario is selected, faculty are encouraged to clearly outline their approach in a class syllabus. As noted, few universal guidelines exist regarding ChatGPT integration. To help instructors think through options, we recommend an exercise to be used in instructor training or with reflective practice groups. For example, instructors could create a hypothetical scenario for this use case and discuss it in class, while also reviewing the institution’s policies on academic integrity.