Reconsidering Literary Excellence: When AI Imitates an Award-winning Poem

Abstract Creativity and diversity have been two of the most noble pursuits in the literary realm. While some claim that local literary competitions encourage such value, some argue that these competitions are usually biased towards certain aesthetic preferences, commonly known as the “award-winning style”, and lead to an ossifying fashion of writing. Artificial intelligence (AI) is used in this paper to evaluate the winning odds of different pieces of Chinese Modern poems in a specific poetry contest and to test whether the awarded poems have a prominent tendency to a certain style. The AI used in this experiment is DeepSeek V3 (“DeepSeek”). Names of judges and full text of a poem (which might or might not be awarded) are provided, and the true results of the competition(s) are not disclosed. It is found that DeepSeek’s evaluation (i) generally matches the true results and (ii) suggests a set of decisive factors for each case, which shed light on the myth of the “award-winning style”. The results reveal the predictability of certain judgements in local poetry competitions and thus suggest (iii) a potential for replicating the winning features. While such an operation is feasible, the value of human creativity and the diversity of the local literary landscape are put into question. It begs a reconsideration of the common notion of “literary excellence” and urges a revaluation of literary diversity beyond the imitation of AI. Keywords: Award-winning Style; Digital Humanities; Modern Poetry; Prize Culture