Nostalgia-driven and AI-empowered: a tripartite efficacy evaluation framework for poetic imagery translation in Chinese design education

Abstract

Within the global movement of cultural revival, the modern translation of poetic imagery (defined as the process of transmuting classical poetic symbols along with their emotional and cultural connotations into modern design elements) has emerged as a critical concern in Chinese design education, presenting a central tension between AI-driven efficiency and cultural depth. This study addresses three structural faults in current translation practices: fragmented symbolic extraction, weakened nostalgic drive, and over-reliance on AI tools. It establishes a tripartite efficacy evaluation framework encompassing emotion, cognition, and market dimensions, as well as a dual-cycle educational model featuring critical and iterative phases. A controlled experiment with 22 second-year product design majors (divided into an AI-assisted group and a traditional group) was conducted over a 4-weeks design psychology course, focusing on war, boudoir, and pastoral poetry themes. Results show that the AI-assisted group outperformed in emotional resonance (4.22 ± 0.38 vs. 3.54 ± 0.47) and market responsiveness (81.3% ± 8.2% vs. 64.1% ± 10.7%), while the traditional group maintained an advantage in cognitive completeness (83.7% ± 5.9% vs. 80.3% ± 5.1%). The dual-cycle model effectively reduced cultural misinterpretation rates in the AI group from 33% to 12%. Meanwhile, this study proposes the “Nostalgia-Congruent AI Guidelines (NCAI-G),” which regulates AI application from three aspects: symbolic fidelity, nostalgia coherence, and user safety. This study provides a reusable educational framework for balancing AI instrumental rationality and cultural value rationality, advancing traditional cultural design education toward quantitative evaluation-driven iteration.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 1 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.