This unusual year for literary prize organisers has taken a further twist. In recent months, the Commonwealth Prize and the Prix Goncourt have had to address allegations that shortlisted writers may have used AI to produce their work. The writers have denied the allegations but the former went on to award the accused writer; the Académie Goncourt removed Thélyson Orélien’s C’était ça ou mourir (It Was Either That or Die) from consideration. Now Africa’s most prestigious literary award, the Caine Prize for African Writing, has found itself embroiled in allegations about the use of AI in a prizewinning story.
Vara bases the AI claim – which Kundizeza strongly denies – largely on Pangram, an AI-detection software
Vauhini Vara, a journalist who has covered AI for years, has claimed in the Atlantic that The Blue Fighter by the Zimbabwean writer Simbarashe Kundizeza, which won the £10,000 Caine Prize, is likely AI-generated. Vara bases this claim – which Kundizeza strongly denies – largely on Pangram, an AI-detection software.
Kundizeza was robust in his response to the allegations: “For the avoidance of doubt, I wrote ‘The Blue Fighter.’ I did not use generative AI to write it. It is entirely my own work,” he wrote on X. The Caine Prize has released a statement saying it is investigating the claim.
The story was inspired by the unpublished poem that serves as its epigraph – “my father’s hands” by Vangile Gantsho, whom Kundizeza met at a residency in Ghana last year. (Pangram considers the epigraph to be genuine.) Kundizeza adds: “I was inspired by this poem to write a story about my family’s experiences in Zimbabwe‚ including my father leaving his accounting profession and the hardships my family faced as a result.”
Vara does not appear to conduct a particularly careful criticism of The Blue Fighter. She notes the “not x, but y” sentences, says the story “is maudlin and full of tired tropes,” points out “the glut of metaphors,” and its “platitudinous aphorisms.” Besides the first, none of these is a particular mark of AI writing; they are signs of bad writing. AI pieces tend to lack a clear plot, they pile up shiny sentences that add up to nothing, and they run on an illusion of metaphor.
Kundizeza’s story, however, has a clear plot: the strongest plot of all five stories shortlisted for the Caine Prize this year. A man begins to nurse expensive fish in his home. He hopes to sell them when they breed. But he is chasing the wind, and his dream causes his family to move to a cheaper house in Ngoni, “a neighbourhood far from the leafy mid-density suburb we’d known.”
His wife bears with him until she can no longer, and one day confronts him with the truth. The dialogue is not especially good. But the wise words dished out make sense contextually: this is a story about keeping hope alive. It was written to inspire.
When the mother sells the fish tanks because there is no electricity (an obvious comment on broken systems in African countries and what they do to dreams), she keeps a small one, which contains the titular fish.
“Everyone should be allowed to keep one dream, Tinashe. Even if it’s a small one,” she says. This had been foreshadowed earlier in the story when the narrator said, “Father set up his smallest tank first. Just one, this time.”
That kind of tacit echoing is a strong proof to me that a human creative imagination is at work.
In the scene where the mother bargains with Shamu, the buyer, Tinashe says, “I watched them dance back and forth. Mother standing her ground while Shamu circled the tanks like a predator. In the end, they settled on twenty-four thousand dollars.”
None of the verbs used is accidental. “Dance” is metaphorical. As such, “standing” in such a dance makes sense. It is also reasonable that the buyer should “circle,” like a bird of prey. It is the language of flight contained in “circle” that finally “settled”– as a bird might settle. There is more going on in this ordinary writing than meets the eye. An AI can’t do it.
Elsewhere, Kundizeza writes of the father, “He found ways to cut corners, fashioning filters from plastics and charcoal, breeding his own food cultures in jars on our windowsill.” The verb “cut” suggests “fashioning” (the mother is a tailor) – it is an imaginative sequence. In the same paragraph, Kundizeza speaks of “the gentle bubbling of the filters [that] became the soundtrack of our new life.” The next paragraph speaks of the township’s “rhythm.” This use of associated words in shifting contexts is how writers work (even if at different levels of skill).
It is also expected that if this is based on actual experiences, the writer may have a sentimental attitude to his material. The narrator says “the walls [of their new house] sweated during the rainy season” (a joke I doubt an AI can make!) and says the walls were thin – twice. It seems to me that this is an imagination trying to keep fidelity with a memory.
There are cliches but almost no pointless metaphors. The “not this, but that” statements also have a different look and are effective. “And anyway, what was my father now?” Tinashe asks. “Not the civil engineer he’d trained to be. Not the small business owner he’d attempted to become. Certainly not someone the government would be eager to claim.”
Because this is a story about failure, it is necessary to sound all those “nots” – it is a parade of unactualised dreams. (Notice the tacit echoings – be, become, claim.)
Vara says that she might be wrong about this accusation. But that seems to be beside the point. The issue, for her, is the “general aversion” of literary prize organisers to AI detectors: they need to get familiar with it instead of eschewing it.
I disagree with Vara. The only reliable tool for determining what is genuine and fake literature is still the critical imagination. A machine cannot be the final arbiter in such matters (although it can serve as a supplement to human judgement). Human judgement may not be fail-safe (editors at reputable journals have been taken in by AI writing), but human errors still teach us something. A machine, whether it is wrong or right, teaches us nothing.
Comments