Visual Reconstruction as Memory Negotiation: An Iterative Generative AI-Mediated Framework for Oral History
arXiv:2608.07507
Abstract
Oral history and community memory are core resources for historical inquiry, yet spatial and material aspects of remembered scenes can be difficult to externalize and compare when they circulate primarily through verbal exchange. This poster proposes an Iterative AI-Assisted Framework for Visual Reconstruction and Memory Negotiation that uses generative AI not to verify memory or produce definitive reconstructions but to create provisional visual 'probes' that support discussion, revision, and comparison. Grounded in oral history and memory studies and informed by digital humanities critiques of visual authority, the workflow proceeds in five stages: (1) narrative elicitation; (2) generative visual prototyping; (3) participant-led iterative revision (human-in-the-loop); (4) multi-narrator comparison and negotiation; and (5) a negotiated reconstruction archive that preserves final images, intermediate iterations, and records of agreement, uncertainty, and disagreement. A vignette illustrates the method through two elders recalling a no-longer-extant building. The poster highlights ethical and methodological constraints, especially the tendency of compelling images to anchor interpretation, and treats outputs as interpretive artifacts rather than evidence. By archiving negotiation traces alongside visual outputs, the approach supports reflexive public-history engagement and responsible documentation in contexts of visual archive scarcity.
Accepted as a poster abstract at the 2026 Conference of the European Association for Digital Humanities (EADH 2026)