A Note From The CAI Co-Chairs
by Todd Essig, PhD and Amy Levy, PsyD
If someone challenged us to stand on one foot to define what the CAI is all about, we’d say, “that’s easy, staying human in the AI age.” Of course, the story is way more complicated. Staying human today is no easy task. Among other issues, it requires assessing the undertow of today’s AI systems and the various ways they invite – and increasingly, push — us to transform basic human experiences, such as falling in love, making art, performing therapy and writing. This month, we’re turning our attention to the challenges of staying human as readers and writers.
Basic questions remain unanswered and often unasked: How should a writer use LLMs, if at all? What duties of ethics and disclosure do writers owe readers? To themselves? What should readers do now that they can detect with remarkable accuracy when a text has been, either in whole or in part, generated by AI?
Did that last question get your attention? It’s been troubling us as some of what we’re asked to read turns out to have probably been AI-generated. How do we know? We’ve been testing the recently released Pangram 4, the next generation of AI detection technology with a remarkably low false positive rate, reported to be on the order of 1 in 24,000 misidentifications. What should a reader do with a text that Pangram 4 identifies as AI-generated?
One conclusion is that the rhetorical signature of the text was AI and not the voice of the writer; the poetics of the text was machine, not human. Pangram 4 does not show how that voicing happened. There are many possibilities. The writer, maybe under deadline or other external or internal pressure, may have taken a short-cut by avoiding the distinctly human challenge of trying to find the right sequence of words to express what they want to say. At the other end of the spectrum, the writer may have used the LLM to substantially alter or even determine what they are trying to say. Not just the rhetorical signature but the content the poetics expresses. Since writing is thinking, they let the LLM do their thinking for them.
But a reader can’t know the difference. Are we encountering a human mind in machine dress? Or are readers suddenly in a relationship with a mechanical, alien AI “mind?” Trying to stay human is no easy task, even standing on both feet.
The CAI’s response includes convening a panel of writers, thinkers, and publishers for the third installment in our CAI Conversations series. We’ll be talking with a poet (Jo Atkins-Potts); a playwright (Andrew Stein who wrote Disruption); a lawyer, AI activist, and writer of The Human Cost (Sean Perryman); a professor and psychoanalyst (Liz Lunbeck from Harvard’s History of Science); a psychoanalytic journal EIC (Greg Rizzolo of JAPA), and a psychoanalytic writer. The recording should be edited and posted on our CAI Conversations YouTube channel before the holidays.
But this is too complicated an issue for a single event. This month we are also organizing an “Editors Summit.” This will be private meeting for psychoanalytic journal editors where they can safely share the challenges to scholarship and publishing they are encountering. We hope this group of talented peers will brainstorm effective solutions. And if they do, we’ll let you know.
