Intelligent Curation of Art Biennales and Exhibitions

 

Sameer Bakshi ¹, Pooja Srishti ², Jaichandran Ravichandran ³, Vasanth Kumar Vadivelu ⁴, Jyoti Saini ⁵ ,

Arpit Arora ⁶

 

¹ Assistant Professor, Department of Visual Communication, Parul Institute of Design, Parul University, Vadodara, Gujarat, India

² Assistant Professor, School of Business Management, Noida International University, India

³ Professor, Department of Computer Science and Engineering, Aarupadai Veedu Institute of Technology, Vinayaka Mission’s Research Foundation (DU), Tamil Nadu, India

⁴ Department of Information Technology, Vishwakarma Institute of Technology, Pune, Maharashtra, 411037, India

⁵ Associate Professor, ISDI - School of Design and Innovation, ATLAS SkillTech University, Mumbai, Maharashtra, India

⁶ Centre of Research Impact and Outcome, Chitkara University, Rajpura- 140417, Punjab, India

 

 

1. Introduction

Historically, the curation of art biennials and exhibition arrangements of large scale exhibitions has been based on the human intuition, cultural research, and aesthetic sense of framing in order to impart the artistic accounts that can capture the audience in a broad manner Zylinska (2023). The rising degree of digitization of art worlds and the geometric rise in the volume of cultural production in the twenty-first century has presented new challenges to curators, institutions and cultural policymakers like never before. Previously only available in a select number of physical spaces, biennials are currently experienced through the network of artists, collectors, and audiences all connected through physical-digital interfaces Zanzotto (2019). This has seen the role of the curator move beyond a conventional taste-keeper to facilitator of dynamic, data-driven and participatory art experiences. This change requires intelligent curation systems AI-driven packages that could be capable of discerning intent in art, streamlining the visual experience of the exhibition space, and engaging audiences on a one-on-one basis. With the advent of artificial intelligence (AI) and machine learning (ML) into the field of cultural informatics, it presents a glimpse of a chance to rethink the conceptualization and organization of art biennials Zhang et al. (2020). Deep learning networks, especially convolutional neural networks (CNNs) and vision transformers (ViTs) have proven to be outstanding image recognizers and stylistic classifiers, and can be trained to recognize works of art based on their medium, technique and era to a certain degree of accuracy autonomously. The algorithms of natural language processing (NLP) can also be used to semantically comprehend artist statements, curatorial texts, and critical essays, as well as enable machines to chart conceptual connections between the works of art and themes. Together with reinforcement learning (RL) of adaptive engagement of space optimization and recommendation systems of audience personalization Barath et al. (2023), these computation methods enable the creation of intelligent curation ecosystems that adapt and enhance, but do not substitute, curatorial knowledge.

It is against this changing background that the idea of intelligent curation continues to go beyond the theory of algorithmic selection. It consists of a cognitive-computational synergy in which human curators work together with AI systems to make sense of cultural data, extract latent patterns Wu (2022) and construct multi-sensory narratives. Through multimodal data, it is possible to discover theme groups, aesthetic proceeds, and socio-political latency that guide the curatorial narrative by visual, textual and behavioral intelligent curation systems. These systems are also of great use when it comes to biennales, wherein the extent and range of participation increases exponentially such that manual curation may prove untenable Shi et al. (2019). With improved visualization and analytics, curators are able to extract meaningful information out of large amounts of data and provide fair representation and contextual integrity across the exhibitions.

Figure 1

Figure 1 Basic Block Architecture of the Intelligent Art Curation Ecosystem

In addition, the increased significance of the audience-centric design in the new era of exhibition engineering requires devices that are capable of shaping and forecasting approached habits. Curation systems can incorporate behavioral analytics and sentiment analysis modules, simulating audience reactions to various setups of an exhibition or a specific theme, and guide the curators to designs that improve immersion and interpretive accessibility Chang (2021). This data-driven method fits into this larger paradigm of cultural analytics, in which the experience of art is both documented and constituted by computational feedback between curators and audiences and computational agents. It is the introduction of AI in curatorial practice, therefore, that signifies a change in the management of display aspects to be flexible and responsive to cultural practices that change in line with audience cognition and their aesthetic preference as depicted in Figure 1. This study seeks to create and test a conceptual and technical model of the intelligent curation of art biennials and exhibitions Picard et al. (2015). It aims to explore how machine learning, semantic, and curatorial heuristic can all be useful in assisting interpretive decision-making, spatial optimization and ethical representation in large scale cultural events. This study helps to bridge the gap between computational creativity and human interpretation by improving the current comprehension of how AI is applicable as a co-curator to enable new forms of culture dialogue and participatory viewing of art Lee et al. (2020). In the end, the proposed framework is intended to reinvent curatorial intelligence as less a mechanized selection process, and more as a collective process of sense-making, in which technology enhances and at most does not weaken the human side of art.

 

2. Conceptual Framework of Intelligent Curation

An intelligent curation has its conceptual base on the combination of cognitive interpretation, computational intelligence, and cultural contextualization. The art ecosystem of modern art and in particular in the biennials and large scale exhibition curation has crossed beyond the process of linear selection to become an interpretative orchestration of meaning. The framework proposed provides the view of intelligent curation as a multilayered network in which data Christofer et al. (2022), algorithms and cognitive processes are in common to achieve greater depth of interpretation, streamlined operation and cultural inclusivity. The combination of the computational approaches and the curatorial thinking not only provides the thematic consistency but also enables the individual audience to experience but ensures ethical and aesthetic integrity Artese and Gagliardi (2022). The Cognitive-Cultural Understanding is the initial dimension, which is the humanistic domain, wherein the curators and cultural theorists formulate ontologies that determine style, theme, emotion and medium as shown in figure 2. Such ontological networks constitute a conceptual body of knowledge that underlies algorithmic thinking about historical, philosophical and socio-political context Kim (2022). This cultural ontology is reflected in the second dimension known as Computational Intelligence and Reasoning in which it is translated into machine-interpretable forms. Reinforcement learning agents improve exhibition design and topic changes to create a feedback-based curatorial feedback system that balances both the coherence and interest Bruseker et al. (2017).

Figure 2

Figure 2 Tri-Layer Conceptual Framework for Intelligent Art Curation

It consists in its essence of three mutually dependent dimensions, namely: (a) Cognitive Cultural Understanding, (b) Computational Intelligence and Reasoning, and (c) Human-Machine Collaboration. Cognitive culture layer is the humanism layer which is a contextualization of the works of art using past, philosophical, and social-political context Weiss (2011). This interpretive part forms the underlying ontology which lays down the curatorial intent and value hierarchies. Under this ontology, the art objects are mapped based on attributes of style, theme, emotion and medium, which constitute knowledge base in conceptual nature. These selected semantic representations are the interpretive base of computational reasoning machines and algorithms of decision-making at the next layer of computation Vanhoe (2016). These cultural ontologies are implemented into machine understandable form in the computational intelligence layer. In this case, machine learning (ML) algorithms are utilized to obtain and process multi-modal-data pictures, input texts, and behavioral responses in order to determine hidden connections among artworks, themes, and reactions of viewers. These multimodal intuitions are also combined through semantic networks and knowledge graphs, which help the machine to reason regarding aesthetic and cultural associations Chen (2016). The learning agents, which are reinforcement learners, maximize the curatorial choices in space layouts and thematic transitions and get better recommendations by comparing their feedbacks with curator and audience responses. Such information intelligence is an adaptability of the curatorial process into an interactive loop of feedback where interpretive form and experience are continuously measured off.

 

3. System Architecture and Design Methodology

The proposed smart architecture of the proposed intelligent curation system of art biennales takes the form of a multilayered computational system that combines the process of acquiring data with the proposed intelligent curation system of art biennial, AI reasoning and platform human-curator interaction. Its transparency, iterative design makes it adaptable and transparent to different curatorial scenarios and its form of a continuous feedback ecosystem, through which multimodal data visual, textual and behavioral streams pass by analysis, reasoning and decision-making processes so that dynamic planning of exhibitions may be achieved. On the base layer, Data Ingestion and Processing Layer gathers and preprocesses data of the museum archives, artist portfolios, and social media with the help of automated crawlers and metadata extractors. Standardization, enhancement, and semantically tagging of data are done in accordance to cultural ontologies like CIDOC CRM, to create a high quality, machine interpretable data. A central part of the AI Curation Engine is the Visual Analysis Unit (CNN + ViT), which appeals to the style and composition of artworks; the semantics of the text provided by the curator is identified by the Semantic Unit (BERT/GPT); and the Affective Analysis Unit identifies how people perceive the work, based on sentiment and gaze data as shown in figure 3. Such outputs are then combined in Multimodal Integration Module through contrastive learning to determine thematic and stylistic relations. Further learning Reinforcement learning is used to optimise exhibition layouts informed by a reward function between aesthetic continuity and viewer interaction. On top of this, the Knowledge Integration and Ontology Layer maps cultural knowledge graphs (through Neo4j) between artworks, artists, and themes, the reasoning of curators, which can be explained and traced.

Figure 3

Figure 3 System Architecture of the Intelligent Curation Platform

The Human -AI Interaction Layer is an interactive dashboard that allows the curator to see insights, customize recommendations, and test virtual layouts trying AR/VR tools. Lastly, both the Output and Deployment Layer transform curated intelligence onto adaptive exhibition designs and bespoke viewing experiences, and real and simulated feedback systems are actively used to continually optimize performance of systems. The Knowledge Integration and Ontology Layer, which is used as a semantic reasoning center, is positioned above the AI engine. In this case, cultures knowledge graphs are created on the basis of relation between artworks, creators, era and themes. The system integrates the use of graph databases (like Neo4j) to visualize and query these networks helping the curators discover invisible connections and cross cultural conversations within the data. Using the ontology-based inference engine, interpretability is greater, and recommendations can be explained, and curatorial choices can be traced. The interpretive metadata is applied on each node and relationship of the knowledge graph, which makes it transparent and ethically accountable. The feedback obtained at real or simulated exhibitions is once again fed into the AI system in a continuous learning process, so the framework constantly changes according to cultural trends and curatorial feedback.

 

4. Algorithmic Workflow and AI Models

The intelligent curation system operationalizes the conceptual and architectural layers into an algorithmic procedure into a series of computer-based steps that converts raw cultural data into curatorial choices. The workflow is shaped like a closed-loop pipeline, with visual, textual and behavioral inputs constantly flowing through, accumulating and assessing in order to assist in macro-level exhibition planning as well as micro-level artwork placement. Each phase uses a set of models of AI that are optimized to the fact that the environment of biennales is multimodal, and are sensitive to curatorial interventions and responses by the audience. The pipeline operates using a set of multimodal feature extraction steps which encode artworks and contextual materials using machine-readable representations. It works as a closed-loop computational loop which processes raw cultural data into curatorial insights by a series of multimodal stages of AI processing. It starts with feature extraction, in which visual data are processed with CNNs and Vision Transformers to extract low-level (color, texture, geometry) and high-level (style, symbolism, subject matter) features, and textual data (essays and reviews) are represented with the help of language models, trained with transformers. At the same time, such behavioral data as the audience dwell time and sentiment responses are modeled into quantitative engagement features.

 

 

Combining these various inputs is done with multimodal fusion and similarity modeling, which aligns the images, texts and behaviors representations into a shared embedding space by contrastive and metric learning. A similarity function S(i,j) is weighted S(i,j) =8.25 S v +.75 S t +.0256 S b ) is used to provide a balanced focus on visual, textual, and behavioral relationships.

 

 

A layout optimizer using RL is then used to sequence and place works of art in virtual or real space by maximizing a reward function which trades thematic consistency, aesthetic transitions, engagement predictions and logistical constraints.

 

 

In order to be ethically transparent and interpretable, explainable AI elements are used to visualize the rationale behind decisions in the form of saliency maps, attention weights, and graph explanations, and rule-based constraints can be applied by curators to make sure that it is inclusive and culturally sensitive. Lastly, the system facilitates human-in-the-loop validation such that curators can engage with AI-generated suggestions with the accept, refine, and reject options, so the system can adjust itself to the curatorial response and institutional values. The second one is the multimodal fusion and similarity modeling, where the heterogeneous features are logically matched in the same embedding space. Semantic compatibility between modalities is imposed with contrastive learning and metric learning methods which ensure that works of art that have similar visual appearances or similar themes are near each other in the embedding space. Similarity combined functionality.

 

 

Visual similarity (Sv), textual similarity (St) and behavioral similarity (Sb), tunable (with weights 3 3 and 3) to reflecting curatorial interests (i.e. giving preference to conceptual solidarity over stylistic homogeneity). Over this combined space, spectral clustering or density-based algorithms (e.g., DBSCAN, HDBSCAN) are used to extrinsic thematic constellations of works of art that may be utilized to create the backstructure of exhibition spaces, pavilions or narrative lines.

 

 

The process of work ends in a validation cycle of the human-in-the-loop. The curatorial dashboard presents AI-generated clusters, layout proposals, and other audience engagement predictions as interactive situations instead of a final decision. Suggestions have the ability to be accepted, edited, or dismissed by curators, annotated as reasoning, and new hypotheses regarding narrative movement or experiencing space can be put forward. Such interactions are recorded and fed back into the learning elements as supervision cues, gradually bringing the system into alignment with the curatorial ideology of the institution with the thematic purpose of the biennale.

 

5. Data Acquisition and Processing Pipeline

The data acquisition and processing pipeline is the bottom layer of an intelligent curation system, and which serves the purpose of ensuring that the multimodal inputs namely the visual, textual, and behavioral data are collected, standardized and integrated into a single framework of analysis. Metadata such as artist, medium, and date are added to the visual data of museum archives, biennale repositories, and websites, like Google Arts and Culture and WikiArt, and textual information in essays, reviews, and artist statements, as well as behavioral information of audience engagement indicators, make up a complete curatorial dataset.

Table 1

Table 1 Summary of AI Models Utilized in the Data Pipeline
Module	Model Type	Core Function	Key Algorithms	Output Representation
Visual Feature Extractor	CNN, Vision Transformer (ViT)	Style and composition recognition	ResNet, ViT-B/16	1024-D Visual Embedding
Textual Semantic Encoder	Transformer-based NLP Model	Concept and sentiment extraction	BERT, GPT, RoBERTa	768-D Text Embedding
Behavioral Pattern Analyzer	Temporal & Statistical Models	Engagement modeling and emotion mapping	LSTM, HMM, k-Means	Behavioral Feature Vector
Fusion Module	Multimodal Embedding Alignment	Unified feature representation	Contrastive Learning, PCA	Shared Latent Space
Validation Engine	Explainable AI (XAI)	Ethical transparency and interpretability	SHAP, Grad-CAM, LIME	Confidence & Attribution Maps

 

Preprocessing also increases the reliability of data by normalizing images, tagging metadata (CIDOC CRM), generating text cleaning and embedding based on NLP models such as BERT, and smoothing behavioral data to be regular. At the integration stage, multimodal fusion is performed in a shared latent space through contrastive learning and dimensionality reduction methods (PCA and UMAP) and attached to the aesthetic, conceptual, and emotional patterns of a semantically defined Cultural Knowledge Graph (CKG).

Table 2

 

Bias detection, provenance verification and ethical oversight are used to validate data and storage is done using hybrid NoSQL-graph databases to provide scalable, real-time access via APIs. This ethical pipeline combines raw and heterogeneous data about art through an integrated, ethically regulated pipeline to create semantically rich, machine-readable data, and lays the foundation of AI-driven intelligent curation on a culturally contextual basis.

 

6. Case Study: AI-Driven Curation of a Contemporary Art Biennale

A case study was performed to test the presented intelligent curation structure as the planning and organization of a mid-scale contemporary art biennale with the support of AI. The case study has shown how multimodal data may be acquired, machine learning may be used to analyze and curator-AI interaction may be used to discover thematic concepts, spatial optimization and predicting audience interest. Through computational reasoning coupled with human interpretive control, the system generated a data-driven but contextually varied curatorial result that reflected the professional interpretation of the curatorial judgment and increased the operational scalability.

 

6.1. Case Context and Dataset Composition

The simulated biennial data involved about 320 artworks of 150 artists in 20 countries, and this was a mix of various mediums, such as painting, installation, digital art, and documentation of the performance. The database incorporated multimodal elements that include visual, textual and behavioral data in order to allow the analysis of artistic expression and audience interaction in a holistic manner.

Table 3

Table 3 Dataset Composition Overview
Data Type	Source/Description	Volume/Size	Purpose
Visual Data	Gallery archives, artist submissions	320 high-resolution images	Artwork classification & visual analysis
Textual Data	Curatorial essays, artist statements, press reviews	~150,000 words	Semantic and thematic modeling
Behavioral Data	Simulated audience engagement metrics	3,200 interaction records	Engagement prediction & emotional analysis

 

Such organized databanks and aspects of ethical design will all make sure that the intelligent curation model works with a semantic breadth, representational proportion, and environmental responsiveness in line with the diversity and inclusivity that now biennials hope to become a symbol of. Images of high quality were used as visual input, derived either through gallery archives or artists submissions and texts were gathered containing curatorial essays, artist statements, as well as press reviews (around 150,000 words). Previous interaction logs that were previously conducted bi-annually were synthesized into behavioral data, with measures being dwell time, engagement rating, and emotional valence. The metadata of medium, year, dimensions, and thematic keywords were also added to each artwork and mapped to the Cultural Knowledge Graph (CKG) in order to build semantic connections. There was fairness and inclusiveness to ensure that they represented the ethics and balanced the representation of the culture present in the regions or styles; therefore, dataset reweighting techniques were implemented.

 

6.2. Thematic Discovery and Clustering Process

The system produced unified features of visual and textual features using multimodal embedding fusion. A comparison-based learning network brought together image and text embeddings which were used in a shared latent space where similarity relationships exhibited aesthetic and conceptual similarity. Spectral clustering and density-based algorithms (HDBSCAN) were then used to determine emergent clusters based on common themes such as Digital Ecology, Memory and Migration and Post-Human Identity. Curators assessed each cluster using the interpretive dashboard, which represented the artwork as a node that was interconnected with others by semantic relationships. The curators were able to expand and contract clusters dynamically and they observed the way conceptual boundaries changed depending on algorithmic thresholds. This interactive contact formed the power of the system as a cognitive reinforcement system, which can bring to light latent curatorial links, which do not exist in hand-based analysis.

 

6.3. Spatial and Layout Optimization

After thematic clustering, the Reinforcement Learning (RL) module emulated the process of laying out an exhibition through a simulated gallery environment with the help of the Unity3D model. The RL agent considered the placement of each art work as an act on a predetermined space grid. R (s,a) was formulated as:

 

 

where (Ct) refers to thematic consistency, (Ae) the possibility to engage an audience, and (Dv) visual diversity. The agent developed the ability to trade off aesthetic continuity and experiential variation using Proximal Policy Optimization (PPO). The 2,000-episode iterative training generated layouts that maximized both cognitively and emotion-pacing, which were better than manually generated layouts when measured by simulated audience satisfaction indices by 18.7 %.

 

6.4. Audience Response Modeling

Behavioral data was inputted into an LSTM-based predictive model in order to evaluate the potential of audience interaction, and the scores of engagement in various layout scenarios were estimated. The model has an average RMSE of 0.072 and F1-score of 0.89 when classifying high-engagement pieces of art. These predictive data were displayed in heat map on top of gallery blueprints allowing the curators to adjust the spaces and lighting setup repeatedly.

 

6.5. Human–AI Collaboration and Evaluation

The last phase of the case study entailed curator-AI co-curation with the interactive dashboard. Qualitative annotations of algorithmic recommendations to approve or disapprove AI generated themes, to modify layout groupings and to label aesthetic subtleties were provided by curators. Transparency was ensured by the explainable AI (XAI) module which showed reasoning paths, attention heatmaps and thematic justifications of each recommendation. Post session surveys indicated that 92 percent of the curators found that the system was more efficient and 81 percent claimed an increased thematic clarity when compared to traditional workflow. Comparison of three curation modes of manual, semi-automated, and AI-assisted revealed the greatest score in curatorial satisfaction and interpretive diversity using the hybrid model. In addition, the introduction of ethical validation constraints also guaranteed equal representation in gender, geography and genre aspects of the exhibition and created a fair and inclusive exhibition narrative.

The case study confirmed the fact that smart curation is not only a tool of computational efficiency but also an innovative partner in interpretive narration. With AI integrated into a system of feedback, ethically controlled, the curators might be able to experiment with the emergent relationships between cultures on a mass scale, without interfering with the artistic integrity. The combination of the multimodal thinking of the system, space adaptability and interpretation helped in making the curating process more democratic and informed leading to a paradigm shift of how future biennales can be conceptualized and experienced.

 

7. Evaluation and Results Analysis

The testing of the suggested intelligent curation system was based on not only quantitative measures of performance, but also qualitative measures of curators, that confirmed the efficiency of the computation and cultural authenticity. The system showed good performances in three key aspects the AI Curation Engine, Reinforcement Learning (RL) Layout Optimizer, and Audience Engagement Predictor that presented the synergy of algorithmic intelligence and human creativity. Visual Recognition Module (CNN + ViT) had an accuracy of 94.2 as compared to the 8-percent increase in Accuracy of the ResNet-50 and the BLEU score of the Textual Understanding Module (BERT-based) was 0.82 and semantic similarity was 0.88. The Multimodal Fusion Network also achieved a Mean Average Precision (mAP) of 0.91, which confirms its successful use of a contrastive learning method. With a PPO agent, The RL Optimizer increased cumulative reward by 21 percent and audience engagement by 17.3 percent, as compared to manual layouts and the LSTM-based Engagement Predictor achieved a F1-score of 0.89 and RMSE of 0.072, which guarantees accuracy in predicting audience response. Interpretive strength Curatorial reviews of the system established the interpretive strength 88% of AI-generated clusters were deemed by the curators as conceptually coherent or higher with the AI revealing hitherto unknown transnational and symbolic connections. Layouts that were optimized by RL were hailed as having more narratives and the interpretable results (attention maps, knowledge graphs) made curators more trusting and interpretative. A comparison of the manual, semi-automated, and AI-assisted workflow (Table 4) showed that intelligent curation took 42% less time to plan, 31% had better thematic coherence, and 28% better audience prediction accuracy with a Curatorial Satisfaction Index of 0.91 and Ethical Representation Balance of 0.93.

Table 4

Table 4 Comparative Evaluation of Curation Approaches
Metric	Manual Curation	Semi-Automated	AI-Assisted Intelligent Curation
Time Efficiency	Baseline	+22%	+42%
Thematic Coherence	Moderate	+18%	+31%
Engagement Prediction Accuracy	N/A	+15%	+28%
Curatorial Satisfaction Index	0.71	0.79	0.91
Ethical Representation Balance	0.82	0.84	0.93

 

This analysis highlights that smart curation systems have a great impact on the extent, interpretability, and inclusivity of art show design. On a quantitative level, multimodal AI model integration allows making accurate clustering, optimizing layouts and predicting engagement. On a qualitative level, the system will help to develop a more comprehensive curatorial discussion, revealing conceptual patterns and providing a clear visual understanding of its logic. Notably, the research confirms that the role of the curator is not weakened but is decontextualized that would elevate the position of the curator to an AI partner and interpretive planner. Intelligent curation provides a trade-off between the objectivity of computation and the culture by synthesizing machine intelligence and human compassion and moral judgment.

 

8. Conclusion and Future Directions

The discussion of smart curation of art biennials as well as exhibition shows how artificial intelligence can transform the parameters of curation practice by combining computational analytics with human imagination. The suggested model that consists of the multimodal data processing, semantic reasoning, reinforcement learning, and human-AI collaboration is an ethical and scalable solution to the problem of exhibition design. It helps curators to go beyond the usual limitations of time, data mass, and subjectivity, and craft coherent and inclusive histories which can be heard across geographical as well as cultural backgrounds. The results of the study are valid in terms of confirming that AI-assisted systems can improve the process of the curators by objective pattern recognition, thematic clustering, and predictive audience modeling. Through explainable AI (XAI) and knowledge graphs, the system can be explained as cultural sensitive to ensure accountability, interpretability, and the ability to refine algorithms, thereby enabling the curator to validate and refine the insight of algorithms. The model is not only efficient in terms of operational performance, it also enhances the conceptual integrity of the exhibition with an analytical rigor and aesthetic and emotional cognition. As shown in the case study, the intelligent curation system enhanced thematic coherence by more than 30 percent and eliminated curatorial planning time, which is almost by half, and proves its usefulness in future large-scale art events. The digital exhibitions could be further extended in terms of the reach and integrity through integration with blockchain-based provenance tracking and immersive XR technologies. Hertz's (2018) study could also be extended in the future by developing hybrid reasoning systems that interoperate between symbolic AI and deep learning to understand abstract concepts like symbolism, emotion, and artistic intent with a finer degree of sophistication. Conclusively, intelligent curation is a paradigm shift in the management of the art exhibition where the human curators and the intelligent systems mutually develop meaning by sharing of reasoning. The intersection of data science, art, and ethics guarantee the fact that technology will develop instead of substituting the curatorial vision. As biennials and museums are going digital, this framework is leading the way to sustainable, inclusive, and interpretively rich cultural experiences, and it has become a new age of curatorial intelligence.

 

CONFLICT OF INTERESTS

None. 

 

ACKNOWLEDGMENTS

None.

 

REFERENCES

Artese, M. T., and Gagliardi, I. (2022). Integrating, Indexing and Querying the Tangible and Intangible Cultural Heritage: The QueryLab Portal. Information, 13, 260. https://doi.org/10.3390/info13060260

Barath, C.-V., Logeswaran, S., Nelson, A., Devaprasanth, M., and Radhika, P. (2023). AI in Art Restoration: A Comprehensive Review of Techniques, Case Studies, Challenges, and Future Directions. International Research Journal of Modern Engineering Technology and Science, 5, 16–21.

Bruseker, G., Carboni, N., and Guillem, A. (2017). Cultural Heritage Data Management: The Role of Formal Ontology and CIDOC CRM. Springer. https://doi.org/10.1007/978-3-319-49356-6

Chang, L. (2021). Review and Prospect of Temperature and Humidity Monitoring for Cultural Property Conservation Environments. Journal of Cultural Heritage Conservation, 55, 47–55.

Christofer, M., Guéville, E., Wrisley, D. J., and Jänicke, S. (2022). A Visual Analytics Framework for Composing a Hierarchical Classification for Medieval Illuminations. arXiv. arXiv:2208.09657

Chen, Y. (2016). 51 Personae: A Project of the 11th Shanghai Biennale, Raqs Media Collective, Power Station of Art.

Kim, B. R. (2022). A Study on the Application of Intelligent Curation System to Manage Cultural Heritage Data. Journal of Korean Cultural Heritage, 29, 115–153.

Lee, J., Yi, J. H., and Kim, S. (2020). Cultural Heritage Design Element Labeling System with Gamification. IEEE Access, 8, 127700–127708. https://doi.org/10.1109/ACCESS.2020.3008844

Picard, D., Gosselin, P.-H., and Gaspard, M.-C. (2015). Challenges in Content-Based Image Indexing of Cultural Heritage Collections. IEEE Signal Processing Magazine, 32, 95–102. https://doi.org/10.1109/MSP.2014.2372315

Shi, K., Su, C., and Lu, Y.-B. (2019). Artificial Intelligence (AI): A Necessary Tool for the Future Development of Museums. Science and Technology of Museums, 23, 29–41.

Vanhoe, R. (2016). Also Space, From Hot to Something Else: How Indonesian Art Initiatives Have Reinvented Networking. Onomatopee.

Weiss, R. (Ed.). (2011). Making Art Global (Part I): The Third Havana Biennial, 70–80. Afterall Books.

Wu, S.-C. (2022). A Case Study of the Application of 5G Technology in Museum Artifact Tours: Experimental Services Using AI and AR Smart Glasses. Museum Quarterly, 36, 111–127.

Zanzotto, M. (2019). Viewpoint: Human-in-the-Loop Artificial Intelligence. Journal of Artificial Intelligence Research, 64, 243–252. https://doi.org/10.1613/jair.1.11348

Zarobell, J. (2021). New Geographies of the Biennial: Networks for the Globalization of Art. GeoJournal. https://doi.org/10.1007/s10708-021-10476-3

Zhang, J., Miao, Y., Zhang, J., and Yu, J. (2020). Inkthetics: A Comprehensive Computational Model for Aesthetic Evaluation of Chinese Ink Paintings. IEEE Access, 8, 225857–225871. https://doi.org/10.1109/ACCESS.2020.3044341

Zylinska, J. (2023). The Perception Machine: Our Photographic Future between the Eye and AI. The MIT Press.