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

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.

1 Introduction

1.1 Research motivation

Based on the practice of Chinese design education, this study focuses on the modern translation of traditional poetic imagery (e.g., visual symbols in Tang poetry, Song ci, and Yuan qu) (Lyu et al., 2024; Xia, 2024). In this context, “poetic imagery translation” denotes a systematic process wherein visual, emotional, and symbolic elements from classical poetry (e.g., Tang poetry, Song ci) are extracted and re-embedded into contemporary design outputs. This process aims to facilitate cultural continuity while fulfilling modern functional and aesthetic requirements. Such imagery carries the spiritual core of Chinese aesthetics and serves as the key content of the “cultural inheritance” module in Chinese design education (Dong et al., 2023; Xia et al., 2024; Yang and Zheng, 2024). Against the dual backdrop of global cultural revival and the deep penetration of AI technology into the design field, Chinese design education is confronting a core contradiction: the imbalance between the efficiency of AI tools and the depth of cultural translation.

From the perspective of technical application, AI tools such as Midjourney have become common aids for design students. Their ability to generate outputs in seconds significantly improves creative efficiency, but simultaneously triggers issues of “stylistic homogeneity” and “cultural superficiality.” For instance, when inputting the classic imagery “withered vines, ancient trees, dusk crows,” AI often simplifies it into a collage of isolated natural scenes, ignoring the “sorrow of a traveler far from home” constructed through the “enumerative parallelism” rhetoric in Ma Zhiyuan’s Tian Jing Sha⋅Autumn Thoughts (Yuan Dynasty). This translation method, which separates cultural symbols from their emotional core, essentially dissolves the Chinese aesthetic tradition of “integration of scene and emotion.” More alarmingly, AI’s random outputs may cause students to abandon in-depth consideration of imagery connotations, falling into the creative trap of “prompt dependency” and weakening the cultivation of cultural critical thinking.

From the perspective of traditional teaching, poetic imagery translation instruction centered on hand-drawing ensures the depth of cultural understanding but struggles to meet the efficiency demands of modern design. Hand-drawn creation requires substantial time for symbol deconstruction and emotional projection–for example, when creating the illustration book Tian Jing Sha⋅Autumn Thoughts, it needs repeatedly refined line density and negative space to convey the lonely artistic conception. This model is inefficient in the design industry that emphasizes rapid iteration. Surveys indicate that most domestic product design teachers report difficulty balancing “cultural depth” and “market response speed” in traditional teaching, resulting in student works that either lack practicality due to overemphasis on cultural expression or become mere symbol collages in pursuit of efficiency (Wang et al., 2023).

This contradiction essentially reflects the imbalance between technical instrumental rationality and cultural value rationality. Without intervention, Chinese design education will fall into the dual dilemma of “efficiency supremacy” or “rigid inheritance.” Therefore, constructing a translation system that balances AI efficiency and cultural depth has become a key proposition for Chinese design education to achieve “technology-empowered cultural inheritance”–this constitutes the core practical motivation of this study.

1.2 Three-dimensional structural faults of “symbolic-emotional-technical”

The systematic issues in poetic imagery translation in current design education can be analyzed through the three-dimensional structural fault framework of “symbolic-emotional-technical”–this framework is proposed by this study for the first time. Most existing literatures focus on problems in a single dimension (e.g., over-reliance on AI in the technical layer or incomplete extraction in the symbolic layer), while this study integrates the three dimensions for the first time to reveal the interactive impacts of faults at each level and their specific consequences for educational practice.

In the Symbolic Layer, the core problem is the fragmentation of imagery extraction. Students or AI tools only separate isolated visual elements from poetic imagery (e.g., “withered vines” and “ancient trees” in “withered vines, ancient trees, dusk crows”) but ignore the semantic connection and overall artistic conception between elements (e.g., “the loneliness of a traveler far from home”), leading to design works becoming mere symbol collages and undermining the cultural integrity of poetic imagery. For example, a student directly transformed elements like “general” and “battlefield” in “Generals die in a hundred battles” into decorative patterns, but failed to convey the core semantics of “tragic sacrifice” in the verse. Eventually, the design was evaluated as “lacking cultural soul” due to the fracture of symbolic semantics.

In the Emotional Layer, the key fault is the weakened nostalgic drive. As emotional carriers of traditional Chinese culture, poetic imagery need to evoke users’ cultural nostalgic memories (e.g., “moon” is associated with “homesickness” and “reunion”), but current translation practices often ignore this demand–AI-generated “moon” imagery is mostly standardized circular light spots, and traditional hand-drawing teaching also fails to systematically guide students to explore the emotional connotation of imagery, resulting in design works being unable to arouse users’ emotional resonance. A statistics on course assignments showed that only 32% of poetic imagery designs could make users associate with the corresponding cultural emotions, and the remaining works were classified as “visual decorations” due to the lack of emotional expression.

In the Technical Layer, the main problem is over-reliance on AI tools, which directly leads to the weakening or even loss of cultural critical thinking. It is necessary to clarify here: cultural critical thinking is a mode of critical thinking applied to cultural practices in the design field, specifically manifested in: examining the historical context and metaphorical logic of cultural symbols [e.g., the “farewell” implication of “willow” stems from the homophonic culture of “willow” (liǔ) and “retain” (liú) in Chinese], identifying cultural value deviations in design translation (e.g., avoiding the entertainment of “war imagery”), rejecting superficial symbol collage and cultural misinterpretation, and ultimately realizing the poetic imagery translation of “both formal similarity and spiritual similarity” (Casteleiro-Pitrez, 2024). When students over-rely on AI tools such as Midjourney to generate imagery combinations, they will gradually give up in-depth thinking about symbolic semantics and emotions, and even encounter ethical issues where AI transforms “Generals die in a hundred battles” into dark-style cosmetics design–this kind of design dissolves the tragic aesthetics of the verse, which is essentially a typical manifestation of the lack of cultural critical thinking (Lin et al., 2023; Neef et al., 2025; Xu et al., 2024).

The above three-dimensional faults further trigger two core contradictions in design teaching: one is Cognitive Conflict–students rely on AI to mass-produce imagery combinations but cannot interpret the metaphorical logic between “moon setting” and “crows crying” in “the moon sets as crows cry” (e.g., “moon setting” implies time, “crows crying” strengthens loneliness); the other is Ethical Conflict–the “decontextualization” of cultural imagery by algorithms leads to designs violating cultural value orientations (e.g., transforming “boudoir” imagery into overly commercialized jewelry, ignoring its aesthetic characteristic of “reserve and gentleness”). The limitation of “valuing skills over evaluation” in traditional teaching (Luo, 2025) further exacerbates these contradictions–when courses only take portfolio completeness as the evaluation criterion, poetic imagery translation gradually becomes a “technical assembly process,” and the cultivation of cultural depth and critical thinking is marginalized.

1.3 Constructing an efficiency evaluation and educational intervention model

Against the three-dimensional structural faults of “symbolic-emotional-technical” and the resulting cognitive and ethical conflicts revealed in Section “1.2 Three-dimensional structural faults of ‘symbolic-emotional-technical,”’ this study proposes three core research objectives, aiming to construct a poetic imagery translation system that balances AI efficiency and cultural depth, and fills the gap of “single evaluation” and “lack of intervention” in current design education.

The primary objective is to construct an “emotion-cognition-market” tripartite efficacy evaluation framework. This framework breaks through the traditional single aesthetic evaluation paradigm and systematically assesses translation effects through quantitative indicators. The second objective is to design a “critical-iterative” dual-cycle educational model as the core path of educational intervention. Through a structured cycle of “deconstruction-prototyping-evaluation-revision,” this model cultivates students’ ability to balance AI instrumental rationality and cultural value rationality. The third objective is to simultaneously propose the “Nostalgia-Congruent AI Guidelines (NCAI-G)” as a supplementary tool for technical layer intervention. Based on the three dimensions of Cultural Deviation Index (CDI), this guideline adds the requirement of “nostalgia coherence” and clarifies three criteria for AI application.

The above three objectives support each other: the tripartite evaluation framework provides an iterative basis for the dual-cycle model (e.g., a low emotional resonance score triggers revisions in the emotional layer), and NCAI-G provides standards for the technical layer intervention of the model, jointly forming a complete “evaluation-intervention-specification” system to target the systematic problems of poetic imagery translation in current design education.

1.4 Innovative value in theory and practice

The innovative value of this study focuses on three core dimensions: theory, methodology, and education, which specifically fills the gaps in existing research on poetic imagery translation while providing a systematic solution to address the “symbolic-emotional-technical” three-dimensional faults and teaching contradictions proposed earlier. In the theoretical dimension, this study integrates nostalgia-driven theory (Mukhopadhyay, 2024; Xue, 2017) into the design translation framework for the first time, proposing the “emotion-symbol-function” coupling principle–existing studies such as Mandour’s (2025) cultural symbol conversion model only focus on the matching of symbol form and function, but ignore the mediating role of emotion (especially cultural nostalgia) in translation. In contrast, this principle clearly states that “symbol extraction must anchor the emotional core, and functional design must carry emotional value”; for example, the translation of the “willow” imagery must simultaneously satisfy the coordination of “visual form (symbol)-farewell nostalgia (emotion)-practical function (e.g., jewelry wearing).” This integration fills the universality gap in the cultural symbol emotional conversion model, enabling the model to cover multi-themed poetic imagery such as war, boudoir, and pastoral.

In the methodological dimension, this study develops a hybrid “emotion-cognition-market” evaluation tool, realizing a paradigm shift from traditional “empirical judgment” to “data-based evidence” (Leavy, 2022). Existing design education evaluations mostly “value skills over evaluation” as noted by Luo (2025), or rely only on a single dimension (e.g., aesthetic scoring). In contrast, this tool quantifies translation efficacy through multi-source data. This hybrid method not only avoids the limitation of pure quantitative evaluation of “ignoring cultural depth” but also makes up for the deficiency of pure qualitative evaluation of “lacking objective standards,” providing a quantifiable and reproducible evaluation path for poetic imagery translation.

In the educational dimension, the reusable teaching module constructed in this study directly addresses the problem of “disconnection between skill transmission and cultural competence cultivation” in design education (Luo, 2025). Existing courses mostly focus on AI tool operation or hand-drawing skills, but fail to systematically cultivate students’ ability to “balance AI instrumental rationality and cultural value rationality.” However, the “critical-iterative” dual-cycle module of this study guides students to establish associative cognition of “technical parameters-cultural expression” through weekly symbol deconstruction workshops and CDI-driven iterative revisions. This practical innovation also responds to the advocacy of Arts-Based Research (ABR) for “integration of teaching and research,” promoting the transformation of educational goals from “skill output” to “cultural translation competence.”

The above three-dimensional innovations support each other: the theoretical “emotion-symbol-function” principle provides a logical basis for the methodological evaluation tool; the methodological quantitative data provides a basis for the iteration of the educational teaching module; finally, a complete innovation chain of “theoretical guidance-methodological support-practical verification” is formed. This not only enriches the theoretical system of design semiotics and cultural translation but also provides an implementable solution for the inheritance of traditional culture in Chinese design education.

2 Related works

2.1 Theoretical foundations of cultural translation in design

2.1.1 Semiotic translation models

Early studies in design semiotics treated cultural symbols as static signifiers, reducing their complexity to formal attributes such as shape, color, and texture. Eco in his seminal work “A Theory of Semiotics” (Eco, 1979), framed cultural translation as a process of extracting “signifier-signified” pairs, emphasizing the replication of visual form over the preservation of contextual resonance. For instance, when translating poetic imagery like “willow” (a classic symbol of farewell in Chinese poetry), these models focused solely on replicating the curved lines of willow leaves in design, ignoring the emotional narrative of parting embedded.

Chandler critiqued this approach in “Semiotics: The Basics” (Chandler, 2022), arguing that it neglects intertextuality–the dynamic interplay between historical context, cultural conventions, and contemporary interpretation. He noted that symbols like “moon” in poetry are not fixed; they shift from “a companion to solitude” to “a witness to reunion,” requiring translations that preserve such fluidity.

Affective semiotics believes that cultural symbols should be able to evoke multisensory and emotional responses (Ware, 1993). For example, translating “云鬓” (“cloud-like hair”) demands more than rendering wavy lines; it requires designing textures (e.g., matte vs. glossy) that trigger tactile memories of softness, thereby connecting visual form to the intimate emotion in poems like “云鬓花颜” (“Her cloud-like hair, her flower-like face,” from Bai Juyi’s The Song of Everlasting Sorrow).

2.1.2 Nostalgia-driven design psychology

Nostalgia Scale revolutionized design psychology by framing nostalgia as a multidimensional construct (Batcho, 2013; Wildschut et al., 2023), distinguishing between personal nostalgia (tied to individual memories) and historical nostalgia (rooted in collective cultural memory).

Within design education, this duality finds expression in divergent translational paradigms. Personal nostalgia frequently compels students to embed autobiographical allusions within poetic imagery–for instance, one student rendered “autumn wind” into a scarf pattern, drawing inspiration from their grandmother’s storytelling beneath autumnal trees. Historical nostalgia, by contrast, centers on activating communal cultural memories: the “dragon boat” imagery from Qu Yuan’s verses, for example, was translated into furniture designs that evoke collective recollections of the Duanwu Festival.

Yet a critical lacuna remains in extant research: while scholarship has quantified nostalgic intensity (e.g., through Likert scales), it has scarcely probed how nostalgia mediates the nexus between poetic abstraction and product functionality. Consider, for example, a teacup adorned with “plum blossom” motifs derived from Wang Anshi’s line “Plums blossom amid the cold”: though such a design may register high on measures of historical nostalgia, if the motif is positioned awkwardly–impeding grip–it fails to reconcile emotional resonance with practical utility. This disjuncture undermines nostalgia’s efficacy as a mechanism for substantive cultural translation.

2.2 Methodological approaches to cultural translation assessment

2.2.1 Quantitative dominance in Chinese design studies

Empirical research in Chinese design education has long prioritized measurable outcomes, reflecting a broader emphasis on technical proficiency in art education. Quantitative metrics dominate assessment frameworks, with two categories prevalent: skill-based indicators (e.g., accuracy of pattern replication, software operation proficiency) and market-driven indicators (e.g., consumer preference surveys, click-through rates in digital platforms).

Literature (Luo, 2025) revealed that most design education studies rely exclusively on such quantitative tools, often reducing cultural translation to a checklist of “completed tasks”–for example, evaluating a student’s translation of “lotus” imagery (from “接天莲叶无穷碧 The lotus leaves stretch endlessly, their green color blending seamlessly with the sky” by Yang Wanli) based solely on how accurately the leaf veins are rendered, rather than whether the design conveys the poem’s vitality.

This over-reliance on quantification has unintended consequences: it encourages students to prioritize technical precision over cultural depth, leading to designs that “look right” but lack emotional or contextual meaning, as the quantitative framework offers no metric for such intangible qualities.

2.2.2 Emergent Arts-Based Research (ABR)

Against the limitations of quantitative hegemony, Arts-Based Research (ABR) has emerged as a critical counterweight, contesting positivist paradigms by foregrounding subjective experience, narrative inquiry, and artistic praxis in evaluative frameworks (Chilton and Leavy, 2014; Leavy, 2022). The heuristic value of visual ethnography is exemplified in analyses of student sketchbooks, which illuminate how learners negotiate poetic ambiguity. For instance, one student’s iterative renderings of “moonlight”–drawing from Li Bai’s line “床前明月光 (The bright moon shines before my bed)”–reveal a notable evolution: from literal “round white circles” to abstract gradients, a trajectory that mirrors their deepening apprehension of the poem’s emotional cadence. Similarly, a/r/tography–a methodology integrating art, research, and pedagogy–has been deployed to document the intersections of students’ personal narratives with poetic texts (LeBlanc and Irwin, 2019).

Yet ABR’s strength in capturing contextual richness is undermined by a significant constraint: it lacks standardized instruments for assessing market viability. A design may resonate profoundly in student documentation–replete with reflections on “homesickness” in Wang Wei’s “A lone stranger in a strange land”–but if it fails to engage broader audiences (e.g., due to overly idiosyncratic symbolism), it risks remaining an academic endeavor rather than a culturally impactful product. This disjuncture from market realities circumscribes ABR’s utility in preparing students for professional design milieus.

2.3 AI’s dual role in cultural translation

2.3.1 Technical accelerators

The integration of AI technologies–encompassing generative algorithms (e.g., Midjourney), style transfer tools (e.g., Neural Style Transfer), and 3D modeling software–has markedly expedited the translation of poetic imagery within design education (Shahzad et al., 2025; Tan and Luhrs, 2024). For instance, style transfer algorithms can rapidly superimpose the ink-wash aesthetic of traditional Chinese painting onto modern product designs, enabling students to materialize visual concepts within minutes rather than days. NLP-driven keyword extraction tools further streamline the process of symbol extraction: parsing lines such as “Iron horses cross the ice river” (from Lu You’s verses), they identify core elements (“iron horses,” “ice river”) and generate associated visual motifs.

Yet this efficiency is accompanied by a critical trade-off: AI’s propensity to flatten cultural complexity risks reducing poetic imagery to decontextualized signifiers. Style transfer may replicate ink-wash textures but fail to capture the “qi” (vital energy) inherent in traditional painting; NLP tools may extract “iron” and “horse” yet overlook the martial ethos embedded in “iron horses.” Consequently, AI-generated translations frequently prioritize formal attributes over essential meaning, echoing the limitations of earlier static semiotic frameworks.

2.3.2 Critical interventions

To mitigate the risks of cultural flattening inherent in AI applications, scholars have formulated evaluative frameworks for gauging the fidelity of AI-generated translations (Li et al., 2024).

This study presents the CDI, a quantitative metric designed to assess the alignment between design outputs and the cultural context of source texts. The CDI quantifies deviation through three dimensions: symbolic similarity, contextual relevance, and cultural appropriateness.

2.4 Academic positioning

This study addresses three critical lacunae in extant literature, situating itself at the intersection of semiotics, nostalgia psychology, and AI ethics (Dang et al., 2025; Obreja et al., 2024):

(1) Theoretical Integration. It integrates semiotic translation models, nostalgia-driven psychological frameworks, and affective design theory into a cohesive “Poetic Translation Triad.” This framework emphasizes dynamic interaction: semiotic extraction–identifying core symbols and their contextual meanings (e.g., “willow” encompassing both its botanical referent and its metaphorical association with farewells–informs nostalgia activation, which triggers personal or collective memories linked to the imagery. This, in turn, underpins affective functionality, ensuring the design elicits the intended emotional response while fulfilling practical utility.

(2) Methodological Innovation. The study proposes a mixed-methods framework that integrates the contextual depth of ABR with quantitative rigor. The ABR arm analyzes student design records through metaphor coding, tracking learners’ negotiation of poetic ambiguity. The quantitative arm of the triad, meanwhile, operationalizes three dimensions of assessment: emotional resonance (as measured by a revised Nostalgia Scale), cognitive completeness (evaluating the extent to which symbols correspond to their poetic meanings), and market responsiveness (gauged through virtual crowdfunding metrics and user interviews). This integration ensures assessments capture both subjective meaning-making processes and objective impact metrics.

(3) AI Ethics Protocol. Building on the CDI, the study formulates the NCAI-G, a three-tiered protocol guiding the responsible deployment of AI. It mandates that AI-generated outputs adhere to three criteria: symbolic fidelity, nostalgia coherence, and user safety (prohibiting the inclusion of traumatic imagery). This framework redresses the technical-affective disjuncture in existing AI tools, ensuring that efficiency is not achieved at the expense of cultural integrity.

The synthesis presented in Table 1 elucidates the study’s academic positioning by addressing key gaps in current research. Specifically, it tackles the emotional void in semiotics, the market disconnection in ABR, and the affective blind spots in AI tools. In doing so, it proposes a more comprehensive approach to the translation of poetic imagery, one that integrates cultural depth, educational rigor, and practical relevance. This foundation supports the methodological and intervention strategies detailed in subsequent sections.

TABLE 1

Table 1. Literature critique synthesis.

3 Materials and methods

3.1 Research subjects and experimental context

This study recruited 22 second-year product design majors (5 male, 17 female, aged 18–22) from a key university in China, all enrolled in a 4-weeks design course. The cohort was selected for its homogeneity in academic background: none had prior formal training in AI-assisted design, though 100% reported basic familiarity with design software, ensuring minimal pre-existing skill bias. The experiment was integrated into the core module on “poetic imagery translation” within the “Design Psychology” course. The tasks were designed around three classic Chinese poetic themes–war, boudoir, and pastoral–which were selected for their diverse emotional tones (tragic, graceful, elegant) and rich symbolic complexity. These themes provided a comprehensive framework to test the participants’ competence in cultural translation.

Twenty-two students were randomly assigned into two groups, with a balanced gender distribution. The AI-assisted group (n = 11) engaged in an AI-integrated workflow, participating in weekly 2-h workshops focused on GenAI prompt engineering. Meanwhile, the traditional group (n = 11) received equivalent training in conventional techniques, specifically sketching with charcoal pencils. Both groups attended 4 h of cultural theory lectures, covering topics such as poetic imagery semiotics and daily life in different dynasties, and were provided with identical materials. Task milestones were established on a weekly basis: initial prototypes were due in Week 2, revised versions in Week 3, and final outputs in Week 4. Progress was monitored through weekly design logs.

3.2 Tripartite assessment framework construction

To systematically evaluate the actual efficacy of poetic imagery translation, this study constructs an “emotion-cognition-market” tripartite quantitative evaluation framework based on the integration of psychological measurement, literary analysis, and market simulation methods. Meanwhile, it designs control measures for the visual style differences between the AI-assisted group and the traditional group, and links it with the NCAI-G to ensure the cultural adaptability of the evaluation.

• Emotional Resonance: measured via a revised Nostalgia Intensity Scale, adapted from Batcho (2013) to include 5 dimensions, 12 Likert-scale items (1 = strongly disagree, 5 = strongly agree) targeting poetic-specific nostalgia (Table 2), responses were collected from 10 independent raters (5 design professionals, 5 literature scholars) to avoid subjectivity.

• Cognitive Completeness (Symbol-Function Mapping Rate): calculated as (Translated Nodes/Total Key Nodes) × 100%, where “key nodes” were annotated by three senior poetry scholars through thematic analysis. A translation scoring 70% would mean 7 of 10 nodes were accurately mapped.

• Market Responsiveness: to validly simulate real-world acceptance within an educational setting, market responsiveness was assessed via a 24-h virtual crowdfunding simulation. This method has been validated in recent design pedagogy research as an effective proxy for assessing market potential and user engagement in educational contexts (Li et al., 2024; Shahzad et al., 2025). The simulation incorporated time-limited campaigns and moderated user comments to mimic real-world crowdfunding dynamics, thereby enhancing ecological validity while maintaining educational feasibility. Metrics included support rate (backer count/total viewers), average pledge amount, and BERT-based sentiment analysis of 100+ user comments (e.g., classifying “feels authentic” as positive, “lacks soul” as negative) to quantify emotional alignment with consumer expectations.

TABLE 2

Table 2. Nostalgia intensity scale: five dimensions and 12 items.

The traditional group do not necessarily need to produce photorealistic renderings, mainly based on the following two considerations: first, to restore the real teaching scenario of Chinese design education–in the basic courses of domestic product design majors, hand-drawn sketches are the core link to cultivate students’ “symbol analysis” and “emotional expression”; introducing rendering software too early will weaken students’ in-depth thinking on cultural symbols; second, to control experimental variables–the core comparison variable of this study is “AI-assisted” vs. “traditional method”; if the traditional group is required to use rendering tools, the interfering variable of “technical proficiency” will be introduced, violating the principle of single variable. To reduce the impact of visual style differences on crowdfunding evaluation, this study clearly labeled on the virtual crowdfunding page that “the design scheme focuses on cultural connotation and functional rationality, and the visual presentation method varies due to different creation methods” to guide evaluators to focus on core indicators.

3.3 Dual-cycle educational model implementation

The “critical-iterative” dual-cycle educational model proposed in this study integrates the NCAI-G and the CDI in depth through a structured process of “deconstruction-prototyping-evaluation-revision.” It specifically addresses the “symbolic-emotional-technical” three-dimensional faults revealed in Section “1.2 Three-dimensional structural faults of ‘symbolic-emotional-technical.”’ The operation and intervention logic of each stage of the model are as follows (Figure 1):

FIGURE 1

Figure 1. Dual-cycle educational model implementation.

• Semiotic Deconstruction Workshops: led by a design anthropologist and a literature professor, these 45-min weekly sessions guided students to unpack imagery layers.

• Prototyping Phases: the experimental group generated 50 initial prototypes via Dreamina AI and selected one for modeling rendering. The control group sketched 5 iterations by hand, refining contours to balance visual simplicity and emotional weight before modeling rendering.

• Tripartite Evaluation & Backtracking: following each prototype submission, the assessment framework generates scores. Outputs with a CDI > 0.4 prompt targeted revisions: for the symbolic layer, add contextual annotations through historical exegesis; for the emotional layer, reinforce the nostalgia anchor; for the technical layer, optimize algorithm parameters.

3.4 Data collection and analysis methods

Data collection spanned quantitative metrics, qualitative insights, and statistical modeling to capture both objective outcomes and subjective processes:

• Quantitative Data: collected weekly, including:

✓ Efficiency metrics: task duration, iteration count and CDI values (as shown in Equation 1 below)

$CDI=1-{{\displaystyle \sum }}_{k=1}^5\left({\mathrm{\omega }}_k\cdot S_k\right)+\mathrm{\lambda }\cdot {\text{Market}}_{\text{Dev}}(1)$

where ω_k: dimension weight; S_k: normalized dimension score (1-5 points); λ: market paradox correction coefficient (usually set λ = 0.2); Market_Dev: market response deviation, ${\text{Market}}_{\text{Dev}}=\left|\frac{{\text{Actual}}_{\text{Support}}-{\mathrm{Predicted}}_{\text{Support}}}{{\text{Predicted}}_{\text{Support}}}\right|$.

✓ Assessment scores: emotional resonance (average rater scores), cognitive completeness (scholar-verified node mapping rates), and market responsiveness (support rates + sentiment scores).

• Qualitative Data: derived from in-depth interviews (20 min per student, semi-structured with prompts like “How did AI influence your understanding of “homesickness”?”) and thematic analysis of design logs.

3.5 Ethical compliance and validity assurance

The study adhered to strict ethical protocols and validity checks to ensure rigor and cultural sensitivity:

• Ethical Compliance. Personal data is anonymized with codes replacing identifiers; AI tools’ application scenarios and limitations are clearly disclosed, with human-AI collaborative outputs labeled; cultural symbols are treated with authenticity, offensive interpretations are avoided.

• Validity Assurance. Reliability: scales showed high internal consistency (Cronbach’s α > 0.78) and test-retest reliability; Double-blind verification: 10% of assessments were re-evaluated by independent raters, confirming consistency; Ecological validity: the simulation mirrored real crowdfunding dynamics (e.g., time-limited campaigns, user comment moderation), ensuring market responsiveness metrics reflected practical viability.

4 Results

By being embedded in the core teaching module of the “Design Psychology” course, a three-stage progressive design of “basic training - thematic practice - comprehensive iteration” is adopted to ensure the systematicity and controllability of data collection. Each round of tasks includes a closed loop of “imagery deconstruction - prototype creation - evaluation and revision - design output,” ensuring that students have sufficient time to deepen their design logic.

4.1 Integrated performance and statistical validation

Table 3 presents the comparison results between the AI-assisted group and the traditional group in the final outputs. The data indicate that the AI-assisted group outperformed the traditional group in terms of emotional resonance, as measured by the revised Nostalgia Scale. Specifically, the AI-assisted group achieved a mean score of 4.22 ± 0.38 (out of 5 points), which represents a significant increase of 19.2% compared to the traditional group’s mean score of 3.54 ± 0.47. This superior performance is largely attributed to the powerful image generation capabilities of AI, which excel at aesthetic identification and the integrity of design.

TABLE 3

Table 3. Comparison of tripartite efficacy evaluation results (AI-assisted group vs. traditional group).

The traditional group maintained a significant advantage in the completeness of symbolic semantics, with an average imagery coverage rate of 83.7% ± 5.9%. This advantage may be attributed to the manual, reflective nature of sketching, which encourages deeper contemplation of symbolic meanings and their functional mappings. Student reflection logs corroborate this interpretation; for instance, one participant noted, “Hand-drawing forced me to think about what each symbol meant in the poem, not just what it looked like.” In contrast, AI-assisted initial outputs tended toward symbolic simplification–a known pitfall of rapid generative processes–though iterative revisions within the dual-cycle model mitigated this gap. It is worth noting that after iterative revisions, the coverage rate of the AI group increased to 80.3% ± 5.1%, narrowing the gap with the traditional group to 3.4 percentage points, which verifies the effectiveness of the dual-cycle model. Student reflection logs supported this interpretation, with one participant noting, “Hand-drawing forced me to think about what each symbol meant in the poem, not just what it looked like.”

Data from the virtual crowdfunding platform showed significant differences: the average support rate of the AI-assisted group’s schemes reached 81.3% ± 8.2%, an increase of 26.8% compared to the traditional group (64.1% ± 10.7%) (p < 0.001). This advantage is mainly due to the accuracy of AI in ergonomic optimization. Further analysis found a significant negative correlation between market response and cultural depth (r = −0.63, p < 0.01), suggesting that cultural translation needs to find a balance between “historical fidelity” and “modern convenience.”

During the 4-weeks experiment, the retrospective revision mechanism was triggered 12 times, with a significant difference in the distribution of themes: war poetry accounted for 58.3% (7/12), boudoir poetry accounted for 25% (3/12), and pastoral poetry accounted for only 16.67% (2/12). This reflects that the symbols of war themes are more sensitive and complex (e.g., “battlefield” imagery is prone to be interpreted in an entertaining way).

Qualitative analysis showed that 73% of students in the AI group recorded the impact of parameter adjustment on cultural depth in their design diaries, indicating that students have gradually established an associated cognition of “technical parameters - cultural expression,” providing subjective evidence for the effectiveness of the educational model.

4.2 Typical case demonstration

This case selects the Yuefu poem “Mulan Poem” from the Northern and Southern Dynasties as the theme for design translation. Starting from the core imagery and situational atmosphere of the poem, it explores suitable product types through reasonable reasoning and creative imagination. The military elements in the poem, such as “buying a fine horse in the east market, a saddle in the west, a bridle in the south, and a whip in the north,” form a striking contrast with the boudoir scene of “By the window, she combs her cloud-like hair, and before the mirror, she pastes yellow flowers.” This contrast ultimately led to the selection of women’s cosmetics as the translation carrier–not only echoing Mulan’s dual identity as “general” and “maiden” but also conveying the poem’s aesthetic of blending strength and softness through product form and function. The comparison of final designs between the AI-assisted group and the traditional group is shown in Table 4.

TABLE 4

Table 4. Comparison of design effects in Mulan poem imagery translation (AI-assisted group vs. traditional group).

5 Discussion and conclusions

5.1 Theoretical and practical contributions

The “emotion-cognition-market” tripartite efficacy model constructed in this study achieves, for the first time, a quantitative evaluation of the design translation of poetic imagery, providing an operable assessment tool for the modern translation of traditional cultural symbols. Its core findings reveal the dual role of AI technology in design education: while it improves design efficiency by 37% and increases market responsiveness by 26.8%, it also clarifies the boundary conditions for technical application– the Cultural Deviation Index (CDI) must be controlled within 0.4; otherwise, it will lead to a significant decline in cognitive completeness (symbol-function mapping rate). This threshold needs to be further tightened to 0.35, especially for culturally sensitive themes such as war poetry.

The dual-cycle educational mechanism proposed in the study demonstrates significant practical value: through supplementary symbolic context, correction of emotional anchors, and algorithm parameter constraints, the cultural misinterpretation rate of the AI group dropped from an initial 33% to 12%, verifying the corrective value of educational intervention in technical empowerment.

Based on the above findings, the study extracts a reusable educational framework: dynamically allocate AI usage rights according to the characteristics of poetic themes (CDI ≤ 0.5 for pastoral poetry and CDI ≤ 0.35 for war poetry); implement a tripartite dynamic assessment of “emotional resonance, cognitive completeness, and market responsiveness”; and establish a pre-emptive cultural review mechanism composed of literary scholars and design ethicists to avoid superficial translation of symbolic collages from the source, which effectively connects theoretical research with teaching practice and provides a specific path for integrating traditional culture and modern technology in design education.

5.2 Limitations and future directions

This study, despite its breakthroughs in theoretical construction and empirical testing, has three limitations. First, the sample of the original study was 22 second-year product design majors (5 males, 17 females, aged 18–22) from a key university. The sample homogeneity is reflected in two aspects: the single educational background (all from the same institution, no cross-school or cross-educational level differences), and the unbalanced gender distribution (females account for 77.3%). It should be noted that this gender ratio is highly consistent with the actual student source structure of Chinese product design majors–female students account for about 75%–80% in product design majors of most domestic design colleges (consistent with the sample ratio of the original study), so the sample has certain practical representativeness in the gender dimension; however, gender differences may potentially interfere with the core variable “emotional resonance.” Second, the reliance on current AI and basic modeling rendering makes it difficult to fully reproduce the textural details of traditional crafts, suggesting a need for future work to integrate culturally fine-tuned AI models or even collaborate with traditional artisans to achieve greater authenticity. Third, while the tripartite framework demonstrates efficacy within Chinese poetic contexts, its applicability to other cultural traditions (e.g., Western sonnets, Japanese haiku) warrants further investigation. Key adaptations would involve recalibrating nostalgia anchors–since nostalgic triggers are culturally specific–and adjusting symbolic mapping protocols to align with distinct poetic structures and aesthetic conventions. Future research should test the framework with multilingual poetry corpora and explore universal principles of cultural translation through cross-cultural comparative studies.

Future research can be expanded in multiple dimensions. At the sample level, we propose collaborating with multiple institutions to collect cross-group data from more than 200 people, controlling variables such as regional culture and AI experience. At the technical level, integrating culturally fine-tuned diffusion models could improve dynamic texture simulation. In tool development, design a real-time monitoring system for cultural sensitivity to automatically identify problematic outputs and provide correction suggestions. In theoretical deepening, a promising direction would be to construct a multilingual poetry corpus to test the framework’s cross-cultural applicability and explore universal principles of cultural translation through international comparative studies.

Data availability statement

The original contributions presented in this study are included in this article/supplementary material, further inquiries can be directed to the corresponding author.

Ethics statement

The studies involving humans were approved by Institutional Review Committee of the School of Arts and Design, Yanshan University. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Author contributions

SH: Writing – original draft, Writing – review & editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This research was funded by 2024 Hebei Province Graduate Professional Degree Teaching Case (Repository) Construction Project of China, grant number KCJSZ2024028.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Abbreviations

ABR, Arts-Based Research; AI, artificial intelligence; NLP, natural language processing; CDI, Cultural Deviation Index; NCAI-G, Nostalgia-Congruent AI Guidelines; GenAI, generative artificial intelligence.

References

Batcho, K. I. (2013). Nostalgia: Retreat or support in difficult times? Am. J. Psychol. 126, 355–367. doi: 10.5406/amerjpsyc.126.3.0355

PubMed Abstract | Crossref Full Text | Google Scholar

Casteleiro-Pitrez, J. (2024). Generative artificial intelligence image tools among future designers: A usability, user experience, and emotional analysis. Digital 4:2. doi: 10.3390/digital4020016

Crossref Full Text | Google Scholar

Chandler, D. (2022). Semiotics: The basics. Milton Park: Routledge.

Google Scholar

Chilton, G., and Leavy, P. (2014). “Arts-based research practice: Merging social research and the creative arts,” in The Oxford Handbook of Qualitative Research, ed. P. Leavy (Oxford: Oxford University Press), doi: 10.1093/oxfordhb/9780199811755.013.003

Crossref Full Text | Google Scholar

Dang, J., Sedikides, C., Wildschut, T., and Liu, L. (2025). Nostalgia encourages exploration and fosters uncertainty in response to AI technology. Br. J. Soc. Psychol. 64:e12843. doi: 10.1111/bjso.12843

PubMed Abstract | Crossref Full Text | Google Scholar

Dong, Y., Zhu, S., Li, W., and Lin, M. (2023). How does the design guidelines for traditional cultural artefacts inspire design in a culturally inspired design process? A comparative study with novice design students. Int. J. Art Design Educ. 42, 49–68. doi: 10.1111/jade.12448

Crossref Full Text | Google Scholar

Eco, U. (1979). A theory of semiotics, Vol. 217. Bloomington: Indiana University Press.

Google Scholar

Leavy, P. (2022). Research design: Quantitative, qualitative, Mixed methods, arts-based, and community-based participatory research approaches, 2nd Edn. New York: Guilford Press, Cop.

Google Scholar

LeBlanc, N., and Irwin, R. L. (2019). “A/r/tography,” in Oxford research encyclopedia of education, ed. G. W. Noblit (Oxford: Oxford University Press).

Google Scholar

Li, H., Zhu, Y., Guo, Q., Wang, J., Shi, M., and Liu, W. (2024). Unveiling consumer satisfaction with AI-generated museum cultural and creative products design: Using importance–performance analysis. Sustainability 16:8203. doi: 10.3390/su16188203

Crossref Full Text | Google Scholar

Lin, A., Paes, L. M., Tanneru, S. H., Srinivas, S., and Lakkaraju, H. (2023). Word-Level explanations for analyzing bias in text-to-image models (No. arXiv:2306.05500). arXiv [Preprint] doi: 10.48550/arXiv.2306.05500

Crossref Full Text | Google Scholar

Luo, N. (2025). A critical review of current trends in art education research in mainland China: Epistemological underpinnings, methodological practices and contexts. Int. J. Art Design Educ. doi: 10.1111/jade.12603 [Epub ahead of print].

Crossref Full Text | Google Scholar

Lyu, Y., Shi, M., Zhang, Y., and Lin, R. (2024). From image to imagination: Exploring the impact of generative AI on cultural translation in jewelry design. Sustainability 16:10065. doi: 10.3390/su16010065

Crossref Full Text | Google Scholar

Mandour, B. A. (2025). Creativity in textile printing design: An integrative framework in design education. Int. J. Technol. Design Educ. 35, 1031–1062. doi: 10.1007/s10798-024-09936-z

Crossref Full Text | Google Scholar

Mukhopadhyay, M. (2024). Nostalgia marketing—A systematic literature review and future directions. J. Mark. Commun. 1–21. doi: 10.1080/13527266.2024.2306551 [Epub ahead of print].

Crossref Full Text | Google Scholar

Neef, N. E., Zabel, S., Papoli, M., and Otto, S. (2025). Drawing the full picture on diverging findings: Adjusting the view on the perception of art created by artificial intelligence. AI Soc. 40, 2859–2879. doi: 10.1007/s00146-024-02020-z

Crossref Full Text | Google Scholar

Obreja, D. M., Rughiniş, R., and Rosner, D. (2024). Mapping the conceptual structure of innovation in artificial intelligence research: A bibliometric analysis and systematic literature review. J. Innov. Knowledge 9:100465. doi: 10.1016/j.jik.2024.100465

Crossref Full Text | Google Scholar

Shahzad, M. F., Xu, S., and Zahid, H. (2025). Exploring the impact of generative AI-based technologies on learning performance through self-efficacy, fairness & ethics, creativity, and trust in higher education. Educ. Information Technol. 30, 3691–3716. doi: 10.1007/s10639-024-12949-9

Crossref Full Text | Google Scholar

Tan, L., and Luhrs, M. (2024). Using Generative AI Midjourney to enhance divergent and convergent thinking in an architect’s creative design process. Design J. 27, 677–699. doi: 10.1080/14606925.2024.2353479

Crossref Full Text | Google Scholar

Wang, J., Xu, L., Yu, X., Lu, H., Huang, Z., Wang, H., et al. (2023). A non-definitive auto-transfer mechanism for arbitrary style transfers. Knowledge Based Syst. 260:110171. doi: 10.1016/j.knosys.2022.110171

Crossref Full Text | Google Scholar

Ware, C. (1993). The foundations of experimental semiotics: A theory of sensory and conventional representation. J Visual Lang. Comput. 4, 91–100. doi: 10.1006/jvlc.1993.1006

Crossref Full Text | Google Scholar

Wildschut, T., Sedikides, C., and Kelley, N. J. (2023). Trait nostalgia: Four scales and a recommendation. Curr. Opin. Psychol. 52:101608. doi: 10.1016/j.copsyc.2023.101608

PubMed Abstract | Crossref Full Text | Google Scholar

Xia, J., Yuizono, T., Wang, T.-Y., and Kim, E. (2024). Intangible cultural heritage course design in the digital age and its effects of interdisciplinary workshop. Educ. Sci. 14:3. doi: 10.3390/educsci14030325

Crossref Full Text | Google Scholar

Xia, S. (2024). Design method based on extensible semantic representation algorithm and its application in product packaging design. Service Oriented Comput. Appl. doi: 10.1007/s11761-024-00430-x [Epub ahead of print].

Crossref Full Text | Google Scholar

Xu, R., Sun, Y., Ren, M., Guo, S., Pan, R., Lin, H., et al. (2024). AI for social science and social science of AI: A survey. Inf. Process. Manag. 61:103665. doi: 10.1016/j.ipm.2024.103665

Crossref Full Text | Google Scholar

Xue, H. (2017). On design and nostalgia: From the perspectives of culture, experience and design strategy. Espoo: Aalto University.

Google Scholar

Yang, X., and Zheng, H. (2024). The present situation, issues and improved schemes of localizing poetry therapy in China. Arts Psychotherapy 87:102116. doi: 10.1016/j.aip.2024.102116

Crossref Full Text | Google Scholar