Digital Sophistry - AI Writing Assistants
The summer of 2022 marked the release of affordable and accessible AI-powered writing assistants, with Microsoft’s decision to deploy and exclusively license OpenAI’s Large Language Model GPT-3. Now, anyone can access one of the most powerful AIs to aid them with writing, digital image creation, and even computer coding. Even before OpenAI announced its open Beta release of GPT-3, over 300 companies had been using the software to generate over four billion words per day. What had initially been designed to be content creation software for the internet has developed remarkably, being able to compose complex responses to questions and it has even been used to write academic essays and compose novels.
Dangers
The dangers and risks inherent in deploying LLM systems like this to the general public are many. LLMs like GPT-3 were not trained in domains of truth—they are essentially stochastic parrots using complex heuristic algorithms to answer questions. There is no logic or reasoning behind an LLM’s response. What’s more, the training data for these LLMs contains all the biases, toxicity, and hatred we’ve become accustomed to seeing in our discourse because LLMs were trained by scraping wide swaths of the internet.
Challenges to Authorship, Academic Honesty, and Attribution
Furthermore, since AI is the author of content it does not copy material but generates it. An AI output cannot be copyrighted. Thus, if a student generated an essay through an AI there would be no plagiarism, as there is no human author. What’s more, there are no reliable means to detect what is organically authored vs. what was produced by an AI. Even if we do somehow arrive at a policy that clearly contextualizes AI generation as cheating because it is incompatible with writing to demonstrate knowledge and build some type of reliable detection device, millions of writers and corporations will have embraced this technology and it will likely be impossible to discern what is and what is not AI-generated material.
Why We Should Embrace AI
As educators, we cannot ignore the possibilities or potential challenges to writing posed by this new technology and what this may mean for student writing and learning. We should embrace AI writing within the context of writing to learn and use this new technology as a tool to help writers further develop their rhetorical skills, critical thinking, and writing ability. Doing so is vital to ensure that tool adoption does not drive further inequity among our students. We may well be at a crossroads in education as we were decades ago when many educators viewed the adoption of the internet as a teaching tool with skepticism. Those who adopt LLMs as part of their writing process will likely be able to complete complex writing and research tasks much quicker than their peers who don’t, and like those who embraced the internet early on, be more prepared to go into job fields that demand they use these new technologies in their careers.
AI as Digital Sophists
One way of viewing this new technology is as a form of twenty-first-century sophistry. The sophists were traveling teachers of rhetoric in ancient Greece who would teach rhetorical skills to the masses for money. They had no interest in seeking truth through their teachings and thus were amoral in their approach to rhetoric. GPT-3 and other LLMs are equally amoral, having no training or grounding in the domains of truth, morality, or ethics. For that, teachers and especially teachers of writing, remain crucial in helping guide students and maturing writers with this technology. Indeed, AI will not replace human-centered teaching, just as the arrival of the internet did not doom universities and teachers, but further establish the need for teachers to help guide their students through emerging and complicated challenges posed by new technologies. The arrival of AI writing—digital sophistry—is no different.
Establishing Standards and Boundaries
Establishing clear ethical standards and boundaries with students matters when dealing with digital sophists. We first need to establish what is and is not allowed when we ask student writers to integrate AI writing within their process.
AI-Assisted Writing vs. AI-Generated Writing
With the rise of AI writing assistants, students must take special care to ensure that they use this new technology ethically and honestly. In our class, we will distinguish between 'AI-assisted writing' verses 'AI-generated writing'. AI-assisted writing is only permitted in this course provided a student uses an AI writing assistant as a collaborative tool to help the student with the development and advancement of their own writing process. Collaborating with an AI writing assistant can include brainstorming, outlining, and drafting, so long as there is substantial writing, research, and composing by the student which is not generated solely by the AI. 'AI-generated writing' means there has been little or no involvement from the student as an author, with the majority of the writing being generated by an AI. The goal of using AI-assisted writing in this class is to help students develop their writing process and critical thinking, not to replace or substitute for either. Therefore, using an AI to generate writing or compositions without substantial original contribution from a student is neither acceptable nor allowed.
Next, we need to inform students to never trust an output from AI at face value. As mentioned previously, these LLMs can provide false, misleading, biased, and toxic responses. Students and teachers must take care in approaching how to use critical thinking in interrogating any AI outputs.
Notice to Students
The AI output may contain offensive, biased material, or material that is false of misleading--review it carefully before using anything suggested by the AI within your academic work.
If you take a suggestion from the AI, you are ultimately the one responsible for using it or discarding it. Thus, I suggest carefully reading and making thoughtful judgments on any suggestions the AI produces before integrating it into your writing.
Rhetorical Heuristics
AI writing assistants are a great example of a digital rhetorical heuristic. A rhetorical heuristic is a strategy for thinking about a problem and is used as a way to help a writer order their thoughts and scaffold their thinking. These rhetorical heuristics have been used by writers and rhetors for centuries and serve as powerful tools to aid one’s critical thinking and provide a roadmap to problem-solving and issue identification. Now, we can combine established rhetorical tools and integrate them within a new one.
A student writer can use the power of an AI writing assistant to examine a topic, issue, or problem in ways that they may not have previously been able to using pen and paper. One example of this is with creating a synthesis matrix within GPT-3 to randomly generate connections and themes on several different topics. Using such a matrix as a teaching tool can help a maturing writing order their thoughts and provide a framework to establish relationships between texts before integrating their own ideas and critical judgments. A teacher might assign a student to generate a simple synthesis matrix, like the one below, then use the output as a guide to create several more categories and connections, while also encouraging them to clarify and deepen the AI’s output using their own writing.
