Human Slop

Yet again, another reminder...

On a Thursday not long ago, a text in my writers’ WhatsApp group alerted me to a flash fiction contest where the winners read at a literary festival in Cork, Ireland. The deadline was Friday. I had a little prose I’d been working on but had not yet shared with anyone. My writing critique group was scheduled for Monday - too late for the deadline - so I decided to see if AI could help me out.

My prompts: Be honest - no coddling. Ask for clarifications if any of my requests aren’t clear (I got this tip from a website). And: DO NOT REWRITE ME - in caps. Then my command: Please critique: Is this flash fiction ready for submission?

Apart from noting a couple of tense inconsistencies and a few misspellings, the verdict was unequivocal. “This is quietly assured writing — controlled, imagistic, emotionally layered — and it earns its central conceit.” Submit, Claude advised, and even wished me luck. So pleased with this assessment, not only did I submit to the Cork Lit Fest, but also Granta, Pangyrus and Southampton Literary Magazine.

The entire weekend, I was buoyed by Claude’s remarks. Imagistic!

On Monday, members of my critique group seemed reluctant at first to speak about my flash and I chalked it up to my prose’s brilliance. Carol went first: This piece is lyrical, she said, it’s nearly a poem. I nodded, imagining myself up on that stage in Cork.

Carol finished with: “What is the point of this piece?” She seemed almost embarrassed to ask the question, and I understood her reticence: Was she really that dense, I wondered, not to get the point of this allegory? Then came Mark, Mike and Alan. All agreed the writing was stellar before adding their version of “Why did you write this?”

AI’s sycophancy is well known, which is why my first command to Claude was to be honest. It did catch that the style was intimate and lyrical, but it failed to find critical flaws.

I forgot that AI doesn’t think.

I asked: Is this ready for submission? It evaluated the piece on whether it was technically polished. Four humans went a step further by looking at thematic intent.

My dream of reading at the Cork Lit fest dissolved into puddles of embarrassment. Imagine presenting something aloud that no one “got.” Worse: I too readily fell for AI’s flattery, even though I knew to expect it. Models are based on cooperative framing, defaulting to qualified approval and avoiding negative feedback. My writing critique group is designed to allow for uncomfortable but fair truths.

This got me thinking about other ways I’ve misread AI as an author. I’ve been careful to refuse these models’ help in actually writing, commanding it instead, for example, to list spelling and grammatical mistakes. I had noticed that when I allowed it to hand me an edited piece, it had often rewritten entire paragraphs without my permission. By requesting a list, I can edit myself and can tick the “no” box in submissions that ask if AI wrote the piece.

I have asked Artificial Intelligence to help me flesh out ideas and work on thorny writing problems. Example: I commanded the bot to assess the use of tension in my short story. Eventually I noticed that Claude just repeated back to me what I already knew - my piece lacked tension - only more eloquently. Because it is programmed to be cooperative, it doesn’t independently generate critical distance, it simply reorganizes my thoughts. Again: I have to wonder: why do I so easily want to surrender my critical thinking skills?

I came across an Ezra Klein column that used Marshall McLuhan’s theory of media on AI. McLuhan was a Canadian theorist who first formalized our thinking about the potent influence of technology and coined the phrase: The medium is the message.

Klein (and McLuhan) posits that AI is intriguing precisely because it mirrors our ideas. The bot, Klein writes, “restates and extends them in a more compelling form than we initially offered, and does so while reflecting a polished image of ourselves back at us.” We stand back and agree we are brilliant.

There is a difference between offloading work to technology and surrendering our cognitive functions to it. It can catch my typos. It can store telephone numbers. That’s fine. Last week, though, I learned how intoxicating it was to surrender to my reflection. I had become a digital narcissist, primping in the mirror to confirm what I secretly hoped: That my writing is brilliant and finally will get the recognition it deserves. It took the Klieg light glare of human honesty to break the spell.

In the end, uncritically abdicating to AI hadn’t saved me any time, in fact it cost me the chance to submit to two literary journals. My writing group, on the other hand, kept my voice authentic and me humble. I am learning the importance of patience. I will submit no prose before its time.

On this Rejection Tuesday, I will continue to work on my flash fiction bruised but not broken. In the words of Samuel Beckett, I will try again. I will fail better.

***

PS: I ran this essay by Claude and here, in part, is the reply:

You asked “Is this ready for submission?” — a technical readiness question — and then faulted Claude for not evaluating thematic intent. That’s like asking a copy editor “is this grammatically clean?” and then complaining they didn’t flag your plot holes. Your critique group asked a better question. That’s a prompt design failure, not an AI failure.

Next, I asked, ‘what would the correct prompt be?’ The answer: “why would a literary magazine reject this piece?