Pedagogical opportunities of AI and writing
Research rationales and specific applications
Undoubtedly, the future of writing will be shaped by generative AI tools such as ChatGPT. A report from Goldman Sachs researchers has projected that AI will lead to “significant disruption” with around “300 million full time jobs” and argue that “two-thirds of current jobs are exposed to some degree of AI automation while generative AI could substitute up to a quarter of current work.” It is crucial in this era of rapid change that domain experts contribute to shaping futures in ways that align with our deep knowledge and situated understandings of stakeholders.
Enacted Development of Writing Skills
Writing as Extended Mind
With respect to writing instruction, generative AI offers opportunities to address longstanding educational challenges. As Matthew Overstreet has argued, working memory is a fundamental constraint for the development of writing skills:
“The limited capacity of human working memory is undoubtedly one of the most important biological constraints a writer faces (see McCutchen, 2000). Indeed, Kellogg (2008) goes as far as to say that working memory capacity is ‘the primary constraint’ on the development of writing skill (2). Given this situation, behavior which works to offload cognitive function is of great value to a writer. It should therefore be a central concern of writing scholars. In particular, in light of writing’s shifting material base, we might look for ways in which emerging technologies allow for increasingly sophisticated acts of cognitive offloading.” (Overstreet, 2022, p. 7-8).
Overstreet’s work points to the way that an “extended mind” theory of writing can help us envision writing processes that focus as much on externalized and enacted thinking and writing processes in specific environments and assemblages of media, technologies, and workflows:
“To shift to an extended rather than brainbound understanding of cognition means paying less attention to what writers think (or say) and more attention to what they do. Fully extended, ‘cognitive writing studies’ comes to signify the study of cognition as enacted in environmental manipulations.” (Overstreet, 2022, p. 6)
Solving for Fluency and Feedback
Furthermore, studies of writing process have argued that the limitations of “working memory” have significant impacts on the learning capacities of novice writers:
“… lack of fluent language generation processes constrains novice writers within short-term working memory capacity, whereas fluent encoding and extensive knowledge allow skilled writers to take advantage of long-term memory resources via long-term working memory” (McCutchen, 2000, p. 13)
Hence, one promising application of an extended mind theory of writing would be to offer support and scaffolding for “fluent language generation” in order to free up cognitive capacity at crucial stages in writing and learning processes.
Another significant constraint on writing skills instruction is the ability to provide learners with situated, responsive, and personalized feedback:
“Time lost between writing and receiving feedback is a significant limitation of the conventional model that can be significantly improved using the digital-forward model. A recurring mantra of this workshop was ‘the more a student writes, the better their writing becomes.’ This being the case, time lost waiting for feedback, which is to say time lost from writing, can be clearly seen as an impediment to student progress in the writing course.” (Garn et al., 2021, p. 11)
This insight emphasizes a focus on technological innovations that enable “micro-feedback” and “mico-learning” cycles to help learners manage attentional focus and cognitive capacity for higher-order skill development:
“This ‘micro-feedback’ model uses smaller, micro-learning chunks that provide more feedback opportunities and mini-scaffolding steps that are easier for under-prepared learners to meet and build a growth mindset.” (Garn et al., 2021, p. 28-29).
Moreover, using academic innovation to help address the learning “bottlenecks'“ of fluency and feedback can help address persistent achievement gaps experienced by students from historically marginalized backgrounds and/or those with disparate access to academic preparation and experiences.
Generative AI and Writing Pedagogy
Generative AI technologies are uniquely suited to helping address these learning challenges. Supporting students in developing skills for using AI to supplement and support their writing processes may allow them to extend their capacities and accelerate their learning. Furthermore, in educational contexts, generative AI may enable applications that support fluency and free up capacity for higher-order learning and may provide scalable ways to provide more responsive and personalized feedback to students at critical moments in the writing and learning process.
Let’s look at some initial ideas for how to take advantage of these opportunities.
Developing Rhetorical Knowledge
An essential aspect of developing writing skills is developing rhetorical knowledge. In fact, this is the first major area of skill development outlined in the WPA Outcomes Statement for First-Year Composition. Developing these skills means students should:
“Learn and use key rhetorical concepts through analyzing and composing a variety of texts …”
“Develop facility in responding to a variety of situations and contexts calling for purposeful shifts in voice, tone, level of formality, design, medium, and/or structure.”
However, a significant impediment to developing rhetorical knowledge is the problem of learning transfer. Previous studies have found that students encounter a bewildering diversity of academic genres and that they experience a disconnect between academic literacies and professional communication (Driscoll and Cui, 2021; Smith et al., 2021). This means that students can struggle to recognize the cross-functional value of rhetorical knowledge and instruction. To counteract this problem, we need ways to help students develop robust and adaptive understandings of rhetorical concepts and genre features. This requires engagement with diverse examples to help better abstract concepts and practice applying them across different contexts:
“Students need many examples when learning complicated concepts (Kirschner & Heal, 2022). When confronted with new and complex ideas, adding many and varied examples helps students better understand them. If students are presented with only one example, they may focus on the superficial details of that example and not get at the deeper concept. Multiple examples of a single concept can help students decontextualize the idea from the example, leading to better recall and understanding.” (Mollick and Mollick, 2023, p. 3)
In traditional writing instruction, providing these types of learning experiences is time and resource intensive. However, AI can assist in this process: “AI can generate numerous examples in very little time.” (Mollick and Mollick, 2023, p. 4).
Suggested Applications
Artificial Intelligence (AI) has the potential to significantly enhance learning activities that help students develop rhetorical knowledge in writing courses. By using AI, educators can create various exercises that emphasize understanding and application of rhetorical and genre strategies. Some ways in which AI can assist in these learning activities include:
Rhetorical Design Repetitions: AI can be used to generate exercises that require students to write messages using the same information but for different audiences, purposes, or media. This would help students recognize and adapt to various rhetorical situations, fostering a deeper understanding of rhetorical concepts such as the rhetorical situation (audience, purpose, message, context) and rhetorical appeals (ethos, pathos, logos).
Genre Strategy Comparison: AI can help create tasks that ask students to write messages for the same rhetorical situation using two distinct genre strategies, and then compare and contrast the effectiveness of each approach. This encourages students to analyze writing critically and better recognize the strengths and weaknesses of specific rhetorical and genre strategies.
Message Competitions: AI can facilitate in-class message case competitions by generating cases themselves and by helping student teams execute a chosen genre strategy from multiple options. The AI system can also assist with evaluating the effectiveness of each team message based on its rhetorical strategy and provide feedback to help students refine their skills. This interactive and engaging approach motivates students to learn and apply rhetorical knowledge effectively and evaluate different genre strategies and approaches in a competitive environment.
Ex. Rhetorical Design Repetitions: “Persuading different audiences to support a community garden initiative.”
Developing Critical Thinking, Reading, and Composing Skills
Hallucinations, confabulations, or “bullshitting” is a widely recognized limitation of current LLM tools such as ChatGPT and Bing AI. However, this bug can become a feature insofar as it can help us address several perennial challenges for students’ development of critical thinking, reading, and composing skills. As WPA Outcomes Statement continues, it identifies these as crucial skills:
“Locate and evaluate (for credibility, sufficiency, accuracy, timeliness, bias and so on) primary and secondary research materials, including journal articles and essays, books, scholarly and professionally established and maintained databases or archives, and informal electronic networks and internet sources
Use strategies—such as interpretation, synthesis, response, critique, and design/redesign—to compose texts that integrate the writer's ideas with those from appropriate sources”
Ryan Watkins has suggested that asking students to create a draft using an AI tool and then fact-check and revise that draft can be an excellent way to help students practice critical thinking skills.
As students learn to engage AI-generated writing with healthy skepticism and recognize the necessity of double-checking the accuracies of both facts and interpretations, this approach can become a habit that they extend to their own writing processes. This could include applications for reading primary and secondary research materials as well as composing their own draft materials. Furthermore, AI can assist as a personalized coach that supports engagement with concepts, frameworks, and enacted skills relevant to critical thinking.
Possible Applications
Develop Research Notes
Discover and Evaluate Sources: Ask students to use AI assistance as a coach and/or tutor for engaging in source evaluation using an established approach such as the SIFT method.
Generate and Evaluate Source Summaries: Ask students to use AI tools to summarize research articles related to their chosen topic. Then, ask them to evaluate the AI-generated summaries, identifying potential gaps or inaccuracies, through further exploration of the original sources.
Organize and Analyze Source Material: Students can use AI to generate “matrix organizers” in the style of the SOAR methodology which has been demonstrated to support learning and argumentative writing (see Kauffman & Kiewra, 2010).
Ex: SOAR Matrix Organizer
Develop Arguments
Create Argument Outlines: Students could use AI to help generate argument outlines and diagrams based on topical prompts or prepared research notes. These outlines could be created and evaluated using different argumentative models (e.g., Toulmin method) and genre strategies (e.g, a elevator pitch).
Explore Counter-arguments and Develop Rebuttals: Students could ask AI to act as a “Socratic opponent” and suggest counterarguments and areas where more evidence may be needed to improve the overall quality of the draft, fostering a deeper understanding of the topic and strengthening their argumentation skills (Sabzalieva & Valentini, 2023).
Draft and Revise Arguments
Create Argument Drafts: Students can use AI to generate a draft based on their own outlines and research notes, providing them with a starting point for their own critical evaluation.
Evaluate and Revise Claims: Students could use AI to identify claims that could benefit from revisions to strengthen specification, clarification, or clearer links to supporting evidence (Skitalinskaya & Wachsmuth, 2023).
Revise for Knowledge Transformation: Students could use AI to identify “knowledge telling” sentences that over-rely on summarizing facts from sources and suggest ways to revise using Bloom’s Taxonomy to make sentences “knowledge transforming.” Previous research has shown that a knowledge-transforming approach to argumentative writing is strongly associated with evaluations of writing quality and a skill that students typically struggle to learn (Raković et al., 2021).
Fact-Check and Evaluate Sources: Students could use AI to help review specific instances of claims and evidence and identify opportunities for fact-checking. Additionally, AI could help students evaluate sources by posing questions about their validity, relevance, and potential bias.
Ex: Evaluate and Revise Claims
Ex: Revise for Knowledge Transformation
By incorporating AI into various stages of the composing process, students can develop critical thinking, reading, and composing skills that will serve them well in their academic and professional careers.
Developing Writing Processes
Perhaps the most significant challenge and learning opportunity AI poses for the development of writing skills is negotiating their impact on writing processes. After all AI-assisted writing pushes the individual writer to work in ways that build on “extended mind” approaches and may actually be similar to collaborative writing with other human beings. Just as collaborative writing requires thoughtful engagement and communication about process mechanics, so will working with AI assistance.
Of particular relevance to this question are these specific process skills highlighted in the WPA Outcomes Statement:
“Develop flexible strategies for reading, drafting, reviewing, collaborating, revising, rewriting, rereading, and editing
Use composing processes and tools as a means to discover and reconsider ideas
Experience the collaborative and social aspects of writing processes …
Adapt composing processes for a variety of technologies and modalities”
One of the greatest opportunities presented by generative AI is to make it easier to invest in explorations of new process techniques that might otherwise seem too time or resource-consuming. It lowers the barrier to entry for diverse types of skill exploration.
Furthermore, these types of explorations can help students develop more flexible and adaptive repertoires of process skills and gain practice enacting them in situated applications. More broadly, generative AI can act as a powerful tool for iterative prototyping that can make it possible for writers to explore multiple different avenues of development, and “to discover and reconsider ideas” at each stage in the writing process.
Possible Applications
Metacognition and Project-Based Learning
“Train your intern”: Following the metaphor suggested by Mollick, students can be asked to analyze and decompose writing tasks in order to create prompts and structured AI-assisted writing processes. This approach will help students develop a more rigorous understanding of the processes, methods, and skills involved in the specific activity. Additionally, it orients us to focus as much on the goals for a process and means for evaluating success as the use the specific techniques. In this sense, it can help student better connect the why’s, how’s, and what’s of a given learning objective following in ways that align with principles for the universal design for learning.
Evaluate different workflows and project plans: Using similar principles, students can be asked to develop workflows and project plans that might integrate and coordinate multiple tools and technologies including both generative AI and other applications. This type of learning experience can help students consider how different tools and processes can be leveraged based on personal strengths and weaknesses, working styles, and task requirements. Following the insights of Lockridge & Ittersum (2020), a key part of effectively using technologies is selectively using a given tool’s features in ways that meet user needs so that the user is driving action rather than having behavior dictated by the tool. Additionally, with the proliferation of generative AI tools across media, it can be extremely valuable for students to think about using sequences of tools. For example, an LLM may be used to generate a textual description of an image that then can be used as a basis for a prompt in an image-generating tool.
Self-reflective journaling: Students can use AI tools to support interactive journaling. Through the use of specific journaling prompts and reflective frameworks (like those suggested in these apps https://www.mindsera.com/ or https://www.rosebud.app/) students can be supported in reflecting on process techniques and/or more broadly in building capacities for self-regulated learning.
Generating Ideas and Prototypes
Exploratorium: Utilizing the capabilities of AI, students can venture into expansive ideation sessions for their academic writing projects (Sharples, 2023). This method aids in nurturing and refining raw ideas, transforming them into robust research questions, theses, and topic concepts. For instance, when exploring the broad subject of "environmental sustainability," students could use AI to brainstorm a range of potential research questions like, "How do urban designs contribute to environmental degradation?" From there, they might refine this into a thesis statement or even generate related claims and transitions to create a coherent argumentative structure.
Prototype and iteration engine!: By leveraging AI's capabilities, students and writers can experiment with a multitude of approaches and iterations for their academic writings, exceeding traditional brainstorming's potential. For instance, a writer wanting to craft a compelling introduction for an essay on "Climate Change Impacts" might brainstorm 3-4 engaging hooks or opening statements. Using generative AI, they can then produce an initial draft for each of these hooks to swiftly evaluate their effectiveness and potential impact.
In another scenario, a writer trying to explain the complex concept of "Quantum Mechanics" might explore various analogies or metaphors, such as "a game of chance" or "waves in the ocean." With the assistance of generative AI, the writer can develop a more detailed and fleshed-out exposition of each analogy or metaphor, allowing them to gauge which resonates more intuitively and effectively with their intended audience.
Exploring Process and Media Techniques
Expert Inspiration: Students could be asked to find a blog post or tutorial that illustrates a technique or specific project with a generative AI tool, then adapt the technique or put your own spin on the project. For example, a student might use an LLM technique for creating new characters to support creative writing or game design. Or, a student might use a video on how to create consistent character visualizations in Midjourney to develop their own gaming avatar.
Multimodal remix: Students could be asked to use generative AI to translate and remix writing across media. For example, students might be asked to convert a script into a storyboard and then into a video. Or, they could create a product description and then use AI to create a design and 3D printed prototype Additional example might include: Examples include:
Taking a research paper's abstract and converting it into a concise, visual infographic that captures its main arguments and evidence.
Transforming an essay on societal impacts of technology into a podcast script, then using AI to generate voice narration, allowing students to explore how auditory arguments differ from written ones.
Drafting a traditional argumentative essay on climate change and using AI tools to supplement the argument with generated charts, graphs, or even simulations. This would demonstrate how visual data can bolster a written argument.
Reimagining a critical analysis of a literary work by creating an AI-generated visual storyboard, reinforcing the student's understanding of the narrative's thematic elements and argumentative interpretations.
Conclusion
This just begins to scratch the surface of what writers, teachers, and learners can do with generative AI. I hope it illustrates some promising starting points for you. If you have questions or idea, let me know in the comments!
References
Driscoll, D. L., & Cui, W. (2021). Visible and Invisible Transfer: A Longitudinal Investigation of Learning to Write and Transfer across Five Years. 32.
Garn, M., Kadel, R., & Lopez, E. S. (2021). ReWriting Writing Establishing Pedagogical and Evidentiary Paradigms for Digital-Forward Instructional in Postsecondary Writing Courses. Association of Public & Land-grant Universities and Every Learner Everywhere. https://www.aplu.org/library/rewriting-writing-establishing-pedagogical-and-evidentiary-paradigms-for-digital-forward-instructional-design-in-postsecondary-writing-courses/file
Kauffman, D. F., & Kiewra, K. A. (2010). What makes a matrix so effective? An empirical test of the relative benefits of signaling, extraction, and localization. Instructional Science, 38(6), 679–705. https://doi.org/10.1007/s11251-009-9095-8
Lockridge, T., & Ittersum, D. V. (2020). Writing Workflows. University of Michigan Press.
McCutchen, D. (2000). Knowledge, Processing, and Working Memory: Implications for a Theory of Writing. Educational Psychologist, 35(1), 13–23. https://doi.org/10.1207/S15326985EP3501_3
Mollick, E. R., & Mollick, L. (2023). Assigning AI: Seven Approaches for Students, with Prompts. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4475995
Overstreet, M. (2022). Writing as extended mind: Recentering cognition, rethinking tool use. Computers and Composition, 63, 102700. https://doi.org/10.1016/j.compcom.2022.102700
Raković, M., Winne, P. H., Marzouk, Z., & Chang, D. (2021). Automatic identification of knowledge‐transforming content in argument essays developed from multiple sources. Journal of Computer Assisted Learning, 37(4), 903–924. https://doi.org/10.1111/jcal.12531
Sharples, M. (2023). Towards social generative AI for education: Theory, practices and ethics (arXiv:2306.10063). arXiv. http://arxiv.org/abs/2306.10063
Skitalinskaya, G., & Wachsmuth, H. (2023). To Revise or Not to Revise: Learning to Detect Improvable Claims for Argumentative Writing Support (arXiv:2305.16799). arXiv. http://arxiv.org/abs/2305.16799
Smith, K. G., Girdharry, K., & Gallagher, C. W. (2021). Writing Transfer, Integration, and the Need for the Long View. 24.