DH 2026

Daejeon, July 27–31

Wed, July 2916:30–18:00S070104
Short Paper

Who narrates the silent past? Human-AI engagement in constructing the digital memory of Sanxingdui

Sijia Lu
Renmin University of China, China, People's Republic of · lusijia_lsj@163.com
Boxin Fan
Renmin University of China, China, People's Republic of · ericfan@ruc.edu.cn
Yunyi Hu
Renmin University of China, China, People's Republic of · huyunyihk@163.com

Brief Abstract

Grounded in post-phenomenology, this study employs digital ethnography to deconstruct human-AI engagement in constructing Sanxingdui’s memory. We analyze the tension in co-creation (Alterity) and immersion in reception (Hermeneutics), revealing how algorithmic agency shifts heritage narratives from archaeological verification to negotiated 'affective truth,' redefining humans as ethical gatekeepers.

Keywords: Human-AI collaboration; Digital memory; AIGC; Post-phenomenology; Non-textual heritage

Abstract

Background

Distinct from civilizations anchored by textual archives, Sanxingdui represents a paradigmatic non-textual heritage of the Bronze Age. It possesses highly stylized material remains yet lacks any written records. This asymmetry between material richness and narrative silence provides a vast interpretive space for Artificial Intelligence to narrate this lost civilization. On Chinese social media platforms such as Xiaohongshu, a proliferation of user-generated posts utilizes AI models like Midjourney and Sora to produce digital restorations of Sanxingdui. This practice elevates AI from a back-end data-processing utility to a front-end memory constructor. Against this backdrop, the historical memory of this voiceless heritage undergoes a critical shift: it is no longer merely discovered, but actively generated.

Research Gap

While research on Participatory Heritage Governance, anchored in the Faro Convention, has rigorously examined the devolution of authority to the heritage communities, discourse surrounding AI in heritage remains largely tethered to instrumentalism. Prevailing studies prioritize technical utility, ranging from data analysis to immersive displays, or static ethical verification of authenticity and bias. This perspective, however, obscures a critical dynamic. It fails to interrogate how AI, operating as a non-human actor endowed with agency, does not merely support but actively reconfigures the ontology of memory through constant negotiation with human users.

Research Purpose

Moving beyond instrumentalism, we deconstruct the dynamics through which human intelligence and AI collaboratively construct the digital memory of Sanxingdui. Rather than assessing the veracity of generated outputs, we interrogate how authenticity is negotiated within the friction between human intent and algorithmic agency. By positioning AI as a co-narrator, we examine how it mediates our cognitive engagement with the past, ultimately revealing the shifting human role from a discoverer of history to an ethical gatekeeper of meaning.

Research Approach

Using multi-sited digital ethnography, we report qualitative findings from a mixed-methods project. Fieldwork (Sep 1-Dec 1, 2025) tracked high-engagement posts and public discussions about AI-generated content related to Sanxingdui on Bilibili, Xiaohongshu, and Douyin. Grounded in Don Ihde’s Post-phenomenology, we construct a tri-phasic analytical framework (see Fig. 1) to interrogate the fluidity of human-machine interaction. We employ the Alterity Relation to examine the tension in co-creation, and the Hermeneutic Relation to analyze the immersion of reception. Crucially, we isolate moments in the Enigma where technological transparency collapses, and public controversy erupts, and use these ruptures as a lens to expose the ethical dilemmas of co-constructing heritage memory with AI.

Fig. 1 Tri-phasic Analytical Framework of Human-AI Memory Co-construction

Preliminary Findings

1. Algorithmic Exoticism in the Alterity Relation. In the human-AI co-creation dynamic, AI manifests strong agency as a “quasi-other,” actively prioritizing the dramatization of historical narratives. Fieldwork data reveals the visible traces of this tension: in a viral Xiaohongshu post, serious tags like #Archaeology are juxtaposed with entertainment markers like #SciFi, and suspenseful rhetoric such as “gods descending.” This suggests that algorithmic weighting amplifies the mysterious and extraterrestrial associations of Sanxingdui while diminishing its historical grounding as a human settlement. Consequently, this friction steers digital memory away from archaeological fact toward a de-contextualized Algorithmic Exoticism.

2. Affective Truth in the Hermeneutic Relation. In the reception phase, where AI outputs are viewed as texts, plausibility consistently supersedes historical authenticity. Taking a video posted on Douyin named “artifacts running on streets” as a case, we observe how AI infuses static bronzes with kinetic vitality, generating a visceral Embodiment. Despite being archaeologically absurd, frequent comments like “goosebumps all over” indicate that this interpretive mode allows audiences to feel the body rather than merely see the object. Here, AI functions as an affective mediator, shifting the transmission of memory from epistemic verification to sensory resonance.

3. The Enigma as Ethical Awakening. When algorithmic deduction oversteps, technological transparency collapses, precipitating what Ihde terms an Enigma. This tension crystallizes in reactions to hyper-sci-fi restorations, exemplified by the highly rated comment: “Do not turn history into fantasy.” Such criticism marks more than disapproval; it signals a disengagement from the immersive experience. This Rupture compels the audience to recognize the content not as the past itself, but as an algorithmic fabrication. Precisely, the emergence of the Enigma activates critical digital literacy, transforming humans from passive viewers into active Gatekeepers of Meaning, charged with defining the ethical boundaries of historical narratives.

Conclusion/Implications

Generative AI functions less as a restorer of the past than as a co-conspirator in Sanxingdui’s contemporary myth-making. Digital memory is thus redefined: no longer a static archive, it emerges as a negotiated artifact, contested by human intent and algorithmic bias within the loop of Alterity, Hermeneutics, and Enigma. This necessitates rethinking Human-AI engagement: in the algorithmic age, connecting with the past requires navigating technical mediation. This mediation presents a paradox: it sparks affective resonance through the embodiment of “living artifacts”, yet risks dissolving historical context into sci-fi spectacles. Thus, the human role must evolve. No longer merely archaeologists unearthing historical truth, we are compelled to act as active gatekeepers of meaning. While embracing technology’s affective thrill, we have to critically engage with the Enigma to prevent heritage memory from becoming subservient to the logic of algorithmic weighting.

Originality/value

Theoretically, we extend the Post-phenomenology to AI-generated historical narratives, framing the tension of co-creation as an Alterity Relation and the immersion of reception as a Hermeneutic Relation. It offers a novel lens for dissecting the friction and flow of mechanisms of human-AI memory co-constructing.

Practically, to leverage AI effectively in heritage activation, GLAM institutions should expand their scope: transcending an exclusive reliance on static archaeological authenticity to embrace a mode of digital storytelling that synthesizes affective truth with ethical awareness.

Contextually, by anchoring the inquiry in the non-Western digital environment and focusing on the textless Sanxingdui civilization, this study challenges Anglophone and logocentric dominance in the digital humanities, revealing how algorithmic hallucinations operate in interpretive voids outside the West.

References
  1. Baek, M. (2024): “Leveraging AI and AR to revitalise Suzhou’s canals: Enhancing sustainable cultural heritage awareness via participatory workshops”, in: Architectural Science Association (ed.): Proceedings of the 57th International Conference of the Architectural Science Association 2024. The Architectural Science Association 728–735.
  2. Bond, L. / Craps, S. / Vermeulen, P. (eds.) (2017): Memory unbound: Tracing the dynamics of memory studies. New York: Berghahn Books.
  3. Colomer, L. (2023): “Exploring participatory heritage governance after the EU Faro Convention”, in: Journal of Cultural Heritage Management and Sustainable Development 13, 4: 856–871.
  4. Duan, Y. (2024): “How did the Sanxingdui become a petal of the ancient Chinese civilization flower?”, in: Journal of Humanities, Arts and Social Science 8, 9: 2180-2186.
  5. Foroughi, M. (2023): Heritage beyond singular narratives: Embracing diversity in participatory heritage planning empowered by artificial intelligence. Ph.D. thesis, Delft University of Technology. TU Delft Repositories.
  6. Ihde, D. (1990): Technology and the lifeworld: From garden to earth. Indiana: Indiana University Press.
  7. Sokka, S. / Badia, F. / Kangas, A. / Donato, F. (2021): “Governance of cultural heritage: Towards participatory approaches”, in: European Journal of Cultural Management and Policy 11, 1: 19–34.
  8. Tiribelli, S. / Pansoni, S. / Frontoni, E. / Giovanola, B. (2024): “Ethics of artificial intelligence for cultural heritage: Opportunities and challenges”, in: IEEE Transactions on Technology and Society 5, 3: 293–305.
  9. Wang, H. / Gao, Z. / Zhang, X. / Du, J. / Xu, Y. / Wang, Z. (2024): “Gamifying cultural heritage: Exploring the potential of immersive virtual exhibitions”, in: Telematics and Informatics Reports 15: Article 100150.
  10. Zhang, C. / Wu, G. (2024): “Sanxingdui and ancient Chinese civilization: In conversation with Zhang Changping”, in: Wu, G. (ed.): The civilization of China and the civilizations of the world. Springer Nature 233–259.