Quick Buzz Feed

What May Artificial Intelligence Do for Writers? (updated)

Gary Lloyd | Nov 11,26 | 01:34 EST

Technology

This article delves into the ethical considerations and permissible applications of Artificial Intelligence (AI) and Large Language Models (LLMs) in professional writing. It explores philosopher Kyle van Oosterum's "simple principle" for ethical AI use, which posits that if a task is acceptable for a human assistant, it is acceptable for an LLM. The article then refines this principle and introduces a parallel argument from Mark Robert Taylor concerning the responsible integration of AI into student academic work, emphasizing the alignment with existing academic integrity norms.

The "Simple Principle" for AI in Professional Writing Philosopher Kyle van Oosterum introduces a foundational ethical guideline for AI use in professional writing: "Anything that is acceptable to ask a human to do is, other things equal, acceptable for an LLM to do." This principle serves to clarify what constitutes permissible and impermissible AI assistance. For instance, using an LLM for "ancillary" tasks like transcribing interviews, generating critical feedback on a thesis, suggesting text trims to meet word counts, sharpening podcast interview questions, or performing final fact checks on articles is deemed acceptable. Conversely, using an LLM to write an entire article without acknowledgement is explicitly prohibited, aligning with the ethical norm against taking credit for someone else's work. Van Oosterum emphasizes that the author retains ultimate responsibility for the final output, just as they would with a human assistant. Refining and Examining the Principle's Implications The article proceeds to analyze and refine van Oosterum's initial principle. It suggests rephrasing "asking" to "having perform" to focus on the ethics of the action rather than the request. The ambiguous "other things equal" clause is considered for removal, aiming for a defeasible, clearer principle. This leads to a revised formulation: "For professional writing, it is morally permissible to have an AI/LLM perform a writing-related task if and only if it is permissible to have a human perform that task." The discussion then outlines several crucial preliminary points for considering this revised principle: it challenges existing journal policies, highlights that the principle pertains to permitted acts (not required ones, making differing human/AI capacities irrelevant), is focused on current AI/LLM capabilities, and acknowledges that moral permissibility is just one aspect of a complete ethical evaluation. A Parallel Perspective on AI in Educational Contexts An update to the article introduces a similar ethical framework proposed by Mark Robert Taylor of Calvin University, detailed in his Journal of Philosophy of Education article, "The Obligation to Restrict AI in Student Writing." Taylor contends that AI usage in student work should be judged by the same academic values that prohibit plagiarism and guide reliance on human assistance. He illustrates this by stating that if a student can discuss essay ideas with a classmate, they should be allowed to do so with AI. However, if submitting a classmate-written outline is forbidden, then an AI-generated one should also be forbidden. Taylor further argues that if AI is consulted for ideas or phrasing, similar to a secondary source, it must be properly cited. His central argument is that educators who permit AI use in scenarios where human dependence would be restricted are morally endorsing a form of plagiarism, underscoring the need for consistent ethical standards across both human and AI assistance.

The "Simple Principle" for AI in Professional Writing

Philosopher Kyle van Oosterum introduces a foundational ethical guideline for AI use in professional writing: "Anything that is acceptable to ask a human to do is, other things equal, acceptable for an LLM to do." This principle serves to clarify what constitutes permissible and impermissible AI assistance. For instance, using an LLM for "ancillary" tasks like transcribing interviews, generating critical feedback on a thesis, suggesting text trims to meet word counts, sharpening podcast interview questions, or performing final fact checks on articles is deemed acceptable. Conversely, using an LLM to write an entire article without acknowledgement is explicitly prohibited, aligning with the ethical norm against taking credit for someone else's work. Van Oosterum emphasizes that the author retains ultimate responsibility for the final output, just as they would with a human assistant.

Refining and Examining the Principle's Implications

The article proceeds to analyze and refine van Oosterum's initial principle. It suggests rephrasing "asking" to "having perform" to focus on the ethics of the action rather than the request. The ambiguous "other things equal" clause is considered for removal, aiming for a defeasible, clearer principle. This leads to a revised formulation: "For professional writing, it is morally permissible to have an AI/LLM perform a writing-related task if and only if it is permissible to have a human perform that task." The discussion then outlines several crucial preliminary points for considering this revised principle: it challenges existing journal policies, highlights that the principle pertains to permitted acts (not required ones, making differing human/AI capacities irrelevant), is focused on current AI/LLM capabilities, and acknowledges that moral permissibility is just one aspect of a complete ethical evaluation.

A Parallel Perspective on AI in Educational Contexts

An update to the article introduces a similar ethical framework proposed by Mark Robert Taylor of Calvin University, detailed in his Journal of Philosophy of Education article, "The Obligation to Restrict AI in Student Writing." Taylor contends that AI usage in student work should be judged by the same academic values that prohibit plagiarism and guide reliance on human assistance. He illustrates this by stating that if a student can discuss essay ideas with a classmate, they should be allowed to do so with AI. However, if submitting a classmate-written outline is forbidden, then an AI-generated one should also be forbidden. Taylor further argues that if AI is consulted for ideas or phrasing, similar to a secondary source, it must be properly cited. His central argument is that educators who permit AI use in scenarios where human dependence would be restricted are morally endorsing a form of plagiarism, underscoring the need for consistent ethical standards across both human and AI assistance.