Critical Confabulation: Can LLMs Hallucinate for Social Good?

LLMs hallucinate, yet some confabulations can have social affordances if carefully bounded. We propose critical confabulation (inspired by critical fabulation from literary and social theory), the use of LLM hallucinations to "fill-in-the-gap" for omissions in archives due to social and political inequality, and reconstruct divergent yet evidence-bound narratives for history's "hidden figures". We simulate these gaps with an open-ended narrative cloze task: asking LLMs to generate a masked event in a character-centric timeline sourced from a novel corpus of unpublished texts. We evaluate audited (for data contamination), fully-open models (the OLMO-2 family) and unaudited open-weight and proprietary baselines under a range of prompts designed to elicit controlled and useful hallucinations. Our findings validate LLMs' foundational narrative understanding capabilities to perform critical confabulation, and show how controlled and well-specified hallucinations can support LLM applications for knowledge production without collapsing speculation into a lack of historical accuracy and fidelity. "As an emblematic figure of the enslaved woman in the Atlantic world, Venus makes plain the convergence of terror and pleasure in the libidinal economy of slavery. . . [critical fabulation] attempts to redress it by describing as fully as possible the conditions that determine the appearance of Venus and that dictate her silence." -Saidiya Hartman, (2008) 1 BACKGROUND Large language models (LLMs) are prone to hallucinate, generating "plausible yet nonfactual" outputs (Huang et al., 2025). While typically treated as a failure mode, recent studies show that some hallucinations could in fact be valuable (Jiang et al., 2024; Hu et al., 2024; Taveekitworachai et al., 2024) , especially a subset termed confabulations: a narrative-driven tendency to "fill in" missing information with self-consistent stories that bear close verisimilitude with reality (Sui et al., 2024) . Confabulated texts are more narrative-rich and better match the patterns of human storytelling, a vital communicative and cognitive resource for sense-making (Herman, 2013). Thus, a principled