: “I’m sorry, Dave, I’m afraid I can’t do that” Part 2

TL;DR

This paper discusses the development of AI tools to detect AI-generated text and explores ethical and practical implications for academic publishing.

Contribution

The paper reviews existing AI text detection tools and raises questions about their reliability and ethical use in scholarly work.

Findings

AI detection tools have been developed and tested for identifying AI-generated text.

The effectiveness of these tools varies, raising concerns about reliability and accuracy.

The paper questions how AI-generated text can be used responsibly in academic and professional writing.

Abstract

In the WYPT session at the Baltimore ACA meeting in 2023, I instigated a discussion on the use/abuse/future of AI generated text in publications (and elsewhere). One of the main take-aways from that talk was the prospect of AI detecting itself: that is, AI applications that can detect with some level of accuracy and precision whether a piece of text was generated by human or AI. Since then, there have been such routines developed and tested. One (of many) report did a fairly thorough analysis of the most common software: (W.H. Walters, “The Effectiveness of Software Designed to Detect AI-Generated Writing: A Comparison of 16 AI Text Detectors” https://doi.org/10.1515/opis-2022-0158 )

This talk will present some conclusions from that report and pose some questions on how to proceed: Can there be safeguards to distinguish AI text reliably? Can these contribute to defining potential legitimate uses of AI-generated text while still protecting copyright and IP? To what extent might we be able to use AI-generated text to prepare publications, reviews, grant proposals, etc.