AI-Enhanced Modular Information Architecture for Cultural Heritage: Designing Cognitive-Efficient and User-Centered Experiences

Abstract

Digital cultural heritage platforms face a dual challenge: preserving rich historical information while engaging an audience with declining attention spans. This paper addresses that challenge by proposing a modular information architecture designed to mitigate cognitive overload in cultural heritage tourism applications. We begin by examining evidence of diminishing sustained attention in digital user experience and its specific ramifications for cultural heritage sites, where dense content can overwhelm users. Grounded in cognitive load theory and principles of user-centered design, we outline a theoretical framework linking mental models, findability, and modular information architecture. We then present a user-centric modeling methodology that elicits visitor mental models and tasks (via card sorting, contextual inquiry, etc.), informing the specification of content components and semantic metadata (leveraging standards like Dublin Core and CIDOC-CRM). A visual framework is introduced that maps user tasks to content components, clusters these into UI components with progressive disclosure, and adapts them into screen instances suited to context, illustrated through a step-by-step walkthrough. Using this framework, we comparatively evaluate personalization and information structuring strategies in three platforms—TripAdvisor, Google Arts and Culture, and Airbnb Experiences—against criteria of cognitive load mitigation and user engagement. We also discuss how this modular architecture provides a structural foundation for human-centered, explainable AI–driven personalization and recommender services in cultural heritage contexts. The analysis reveals gaps in current designs (e.g., overwhelming content or passive user roles) and highlights best practices (such as tailored recommendations and progressive reveal of details). We conclude with implications for designing cultural heritage experiences that are cognitively accessible yet richly informative, summarizing contributions and suggesting future research in cultural UX, component-based design, and adaptive content delivery.

1. Introduction: Problem Definition and the Attention Crisis in User Experience

Digital users increasingly engage in fragmented, short-duration interactions shaped by mobile-first habits, pervasive notifications, and multitasking. Empirical findings show that over half of web visits last under ten seconds [1], while work-related screen tasks now average under one minute before an attention shift [2]. Cognitive Load Theory (CLT) reinforces that limited working memory is easily disrupted by extraneous information [3]. Recent HCI studies further highlight micro-session behaviors and rapid task switching as dominant patterns in digital environments. Together, these trends indicate that interfaces requiring sustained, uninterrupted attention face structural engagement barriers [4,5].

1.1. Attention Fragmentation and Its Implications for Cultural Heritage

Cultural heritage platforms—museum apps, virtual tours, and online collections—traditionally rely on dense interpretive content, long narratives, and hierarchical knowledge structures. These characteristics place high demands on users’ cognitive resources, particularly when explored on mobile devices. Studies confirm that users often skim or overlook long-form cultural content, rarely revisit previous screens, and disengage when interfaces present too much detail at once [6,7]. In this context, ‘scaffolding’ refers to providing temporary support to users until they can navigate content independently, and ‘progressive segmentation’ means breaking complex information into manageable segments that are revealed gradually. Without such measures, heritage material can impose both intrinsic and extraneous cognitive load, causing visitors to miss key interpretive information.

1.2. From General UX to Cultural UX

Cultural User eXperience (UX) differs from conventional digital UX because it involves meaning-making, contextual interpretation, and the integration of historical, spatial, and symbolic information [8]. Users often lack prior domain knowledge, making intuitive navigation and clear conceptual pathways essential. When system structures diverge from users’ mental models—e.g., taxonomy-driven menus, unfamiliar categories, or fragmented content—cognitive effort increases and comprehension decreases [9,10]. Prior research shows that cultural interfaces frequently exceed everyday cognitive demands due to dense, non-linear content and heterogeneous media formats [11].

For the purposes of this study, we define ‘Cultural Heritage Tourism’ as encompassing both traditional museum-based experiences and broader urban and digital explorations of cultural assets. This hybrid scope reflects the evolving nature of cultural engagement, where users increasingly navigate between physical sites, online collections, and experiential platforms. Our framework is designed to accommodate this diversity by remaining agnostic to content type and context, ensuring applicability across museum artifacts, city landmarks, and thematic cultural narratives.

1.3. Consequences for Information Architecture (IA)

These attention and cognitive constraints directly affect Information Architecture (IA), the structural design of information environments that governs how content is organized and accessed. Overly dense or poorly structured content can lead to disorientation, information avoidance, and loss of narrative coherence [12]. Findability suffers when navigation does not reflect user mental models or when menu depth and labeling do not align with intuitive conceptual structures [13,14]. Effective IA must therefore balance informational richness with cognitive accessibility, providing modular, digestible units that support exploration without overwhelming the user.

To address these challenges, this paper proposes a modular information architecture explicitly designed to reduce cognitive overload and support progressive disclosure in cultural heritage applications. Beyond UX and content structuring, we argue that modular IA can act as an enabling layer for human-centered AI, supporting interpretable recommendations and adaptive cultural journeys.

This paper makes three main contributions:

(1) Synthesizes Cognitive Load Theory, mental models, and modular information architecture into a theory-driven framework for cultural UX;

(2) Proposes a user-centered, component-based modeling methodology and layered visual framework that connect semantic content units to UI patterns and screen instances;

(3) Illustrates the framework through a comparative analysis of major cultural heritage platforms and a persona-based simulation of adaptive cultural journeys, highlighting implications for human-centered AI and explainable personalization.

To sharpen the focus of this study, we explicitly formulate the following Research Questions (RQs) to guide our inquiry:

• RQ1: How can Cognitive Load Theory and modular information architecture be synthesized to reduce cognitive barriers in digital cultural heritage experiences?
• RQ2: To what extent do current leading cultural heritage platforms support modularity and cognitive efficiency in their user experience?
• RQ3: How can a user-centered, component-based framework facilitate adaptive and personalized cultural heritage journeys?

To guide the reader through the structure of the paper, we provide a brief roadmap of the sections that follow. Section 2 establishes the theoretical foundations of our framework, synthesizing Cognitive Load Theory, mental models, and modular information architecture principles. Section 3 details the user-centered component modeling methodology and introduces the layered visual framework that connects semantic content to interface design. Section 4 presents a comparative structural analysis of three major cultural heritage platforms, evaluating their alignment with our proposed framework. Section 5 demonstrates the framework’s adaptability through a persona-based simulation of cultural UX journeys. Section 6 summarizes the key contributions of the study, and Section 7 discusses limitations and outlines directions for future research in cultural UX and modular design.

2. Theoretical Foundations

The design of cognitively efficient cultural heritage interfaces requires a solid theoretical grounding that connects human cognitive limitations, user expectations, and modular content structures. This section synthesizes three complementary bodies of knowledge—Cognitive Load Theory, mental models and findability, and modular information architecture—to establish the conceptual basis for the proposed framework. Together, these theories provide the principles that guide both the structuring of cultural content and the design of user-centered interaction patterns.

2.1. Cognitive Load Theory in Cultural Interfaces

Cognitive Load Theory (CLT), which posits that human working memory has limited capacity and is easily disrupted by extraneous information, offers a vital framework for designing user-centered cultural heritage interfaces. CLT distinguishes three types of cognitive load: intrinsic (content complexity), extraneous (presentation-related burden), and germane (effort contributing to learning) [3]. Heritage content often carries a high intrinsic load due to its complexity—historical dates, foreign terms, and layered narratives—making it essential to minimize extraneous load and manage germane load to support user comprehension.

Recent studies apply CLT to cultural interfaces. Zhou et al. [15] demonstrated that segmenting heritage content into granular, progressive units improved short-term recall (84.7% vs. 64.6%) and long-term retention. Their CLT-informed serious game design staged information to avoid overwhelming working memory, validating the segmenting principle—akin to progressive disclosure in UX. Further, studies such as Marty [7] and Hacioglu [4] have emphasized the challenges users face in digital museum environments, including fragmented attention and difficulty navigating dense interpretive content. These findings align with our argument that cultural heritage platforms often exceed everyday cognitive demands. Despite these insights, few systems have operationalized Cognitive Load Theory in their architectural design. Our framework addresses this gap by integrating CLT principles—such as progressive segmentation and germane load optimization—into a modular, user-centered structure that supports both comprehension and engagement.

The ‘split-attention effect’—a core concept in Cognitive Load Theory—refers to the increased extraneous cognitive burden that occurs when users must mentally integrate spatially separated information sources (for example, a diagram and its descriptive text) [3,16]. Designers can mitigate this by integrating media and text within unified UI components, using cues like highlights or hover effects to guide attention. Our modular architecture supports this by embedding logically related content within single components, reducing cognitive fragmentation.

CLT also emphasizes germane load, the mental effort that fosters schema construction. Effective interfaces channel user effort toward understanding, not interface navigation. For example, a simple quiz or guided animation can enhance learning without adding extraneous load. Chen et al. [17] found that stepwise animations improved comprehension of intangible heritage crafts by increasing germane load while keeping extraneous load low.

In summary, CLT informs key design strategies for cultural UX: reduce extraneous load, structure content to manage intrinsic load, and foster germane load through meaningful interaction. These principles underpin our modular architecture, which organizes content into self-contained components and reveals them progressively. The goal is to align interface design with cognitive limitations, enabling users to engage deeply with heritage material without feeling overwhelmed.

2.2. Mental Models and Findability

Effective findability in cultural heritage apps depends on aligning the interface with users’ mental models—their internal expectations about how content is organized. These models are shaped by general web habits (e.g., expecting search bars or menus) and domain-specific assumptions (e.g., grouping artifacts by era or type). When an app’s structure diverges from these expectations, users experience disorientation and cognitive strain. Conversely, supporting mental models—through familiar labels, intuitive navigation, or subtle onboarding—enhances usability and engagement [10].

Users form mental models quickly, often from first impressions. For example, seeing categories like “By Era” or “By Location” suggests a taxonomy-based structure. If content is instead organized by catalog IDs, the mismatch leads to frustration. In heritage contexts, this means using plain language (e.g., “Renaissance (1400–1600)” instead of technical terms) and offering multiple navigation paths to accommodate varied conceptualizations—chronological, thematic, or geographic.

Findability also involves information scent: users follow cues that suggest where desired content lies. Pirolli’s Information Foraging Theory posits that users abandon paths when cues weaken or effort increases [13]. If expected tools like map views or timelines are absent or buried, users may disengage. Konstantakis et al. found that users prefer navigation options aligned with their personal exploration styles, highlighting the need for flexible IA that supports multiple mental models [11].

User research is essential to uncover these models. Techniques like open card sorting reveal how users group content, informing component and menu structures. For instance, users might associate “architecture”, “painting”, and “music” under “Renaissance Culture”, rather than by medium. Designing IA around these groupings improves findability and reduces cognitive effort [18,19].

In our modular architecture, mental models guide component definitions and relationships. Using intuitive labels and icons (e.g., map pins for location, timelines for chronology) supports recognition over recall. Consistent UI behavior—such as predictable detail panels—reinforces users’ understanding of system structure.

In conclusion, mental models bridge users and system architecture. Aligning IA with these models—or gently guiding users toward new ones—enhances findability, reduces cognitive load, and enables deeper engagement with heritage content. Section 3 details how user research informs this alignment.

2.3. Modular Information Architecture Principles

Unlike traditional hierarchical sitemaps or fixed taxonomies, modular information architecture (IA) structures content as discrete, recombinable units governed by consistent rules. Drawing from software engineering principles—such as modularity and separation of concerns—this approach emphasizes reusability, flexibility, and scalability [20].

A core principle is context-independent content: each module should be self-contained and meaningful on its own. For example, an artifact module should include essential details (title, image, description) without relying on prior content. This supports progressive disclosure and personalization, allowing users to access relevant information incrementally. Many researchers advocate for modular “chunks” that can be reused across contexts—e.g., the same artifact module appearing in both a map view and a thematic tour [10,21].

Another key principle is the explicit definition of relationships and metadata. Modular IA benefits from structured content models or ontologies (e.g., Dublin Core, CIDOC-CRM; see Section 3.2), enabling dynamic assembly of related modules. For instance, viewing an artist module can trigger retrieval of linked artifact modules via a “created_by” relationship. Patel (2005) highlights how robust metadata enables seamless content repurposing and exploration, avoiding isolated silos [14].

Modular IA also aligns with progressive enhancement, presenting essential modules first and layering additional detail as needed. This mirrors Object-Oriented UX (OOUX), which treats content as reusable objects with defined boundaries and relationships. Though still emerging in the academic literature, OOUX builds on established design principles—such as Sanchez and Mahoney’s work on modular product architectures—emphasizing adaptability and iterative improvement. This modular approach to information architecture aligns with constructivist learning theory and schema theory in education, treating each content module as a ‘building block’ of knowledge that users can independently grasp and then combine—much as learners construct knowledge from discrete, meaningful units [22,23].

Scalability is another advantage. As content grows, modular IA allows new modules and relationships to be added without restructuring the entire system. This view resonates with pervasive information architecture approaches that treat user journeys as inherently cross-channel and cross-context [24]. Europeana’s shift to a linked data model exemplifies this, enabling integration of diverse museum content while maintaining coherence [25].

In essence, modular IA balances coherence and flexibility. Each component is semantically structured and metadata-rich, while the system dynamically assembles content to match user context. This supports non-linear exploration, serendipitous discovery, and personalization, allowing users to engage deeply without cognitive overload.

To summarize, our framework is guided by five principles:

• Componentization of content;
• Context-independence of modules;
• Explicit relationships via metadata;
• Progressive enhancement through layering;
• Adaptability across contexts and user needs.

These principles form the foundation for the user-centered methodology described in Section 3.

3. Methodology: User-Centered Component Modeling and Evaluation Design

Having established the theoretical underpinnings, this section translates those principles into a concrete methodological approach. Building on the theoretical foundations of cognitive load, mental models, and modular information architecture, it describes how user-centric content components are translated into interface structures. The goal is to connect user tasks to semantic content units and then to visual representations through a systematic, layered modeling process, ensuring that each layer of the design supports clarity, findability, and low cognitive load.

Our approach adopts a layered framework consisting of (i) semantic content components, (ii) UI components that present these units with progressive disclosure, and (iii) screen instances that adapt to user context and device constraints. The framework operates conceptually as a model-driven pipeline rather than an implementation method.

3.1. Mapping User Tasks to Semantic Components

The first step is aligning user tasks with semantic content units. Each task (e.g., “explore an artifact”, “compare two experiences”, “plan a route”) must map to one or more content components that contain the needed semantic information. This guarantees that the system supports real user goals rather than internal institutional structures.

This mapping draws on goal-directed and scenario-based design principles, where system objects are defined according to how users conceptualize the domain. For example, a user task such as “learn about the creator of this artifact” requires a Person component linked semantically to an Artifact. Similarly, “browse by theme” requires Theme components with explicit relationships to artifacts, places, or events.

These relationships reflect users’ intuitive conceptual pathways. Users naturally understand connections such as “artifact–creator”, “site–period”, or “event–location”. Structuring components according to these semantic links reduces cognitive effort and improves findability. This is consistent with knowledge organization theory and cognitive semantics, where comprehension is improved when information mirrors real-world conceptual structures.

The result is a user-centered semantic model that establishes which components exist, which attributes they must contain, and how they relate to each other, as shown in Figure 1.

3.2. UI Components and Progressive Disclosure

UI components are reusable interface patterns that present one or more content components. Examples include cards, previews, detail panels, map pins, or timeline elements. Each UI component represents a consistent method of revealing content progressively, beginning with essential cues and expanding into detail on demand [26].

Progressive disclosure is a key strategy for managing cognitive load [3]. It provides users with small, meaningful units of information first—such as an image, title, and brief summary—and allows deeper exploration without overwhelming working memory. This aligns with Shneiderman’s mantra (“overview first, zoom and filter, then details-on-demand”) [27] and reduces extraneous load by limiting unnecessary detail in early interactions.

The design of UI components leverages recognition rather than recall: consistent patterns (e.g., similar card layouts) make interfaces predictable, reducing cognitive friction [9]. Multimedia principles such as spatial contiguity guide visual presentation, avoiding split-attention effects by integrating images and text within the same visual unit.

By decoupling content semantics from presentation, UI components can represent the same underlying component at different levels of granularity—for example, an Artifact component may appear as a thumbnail card, an expanded card, or a full detail view.

3.3. Screen Instances and Contextual Adaptation

Screen instances are full compositions of UI components arranged to satisfy specific user goals and contextual constraints. Unlike static page templates, they adapt based on user expertise, device type, location, or situational needs. This draws on concepts from adaptive hypermedia and context-aware systems [28].

Contextual adaptation supports different modes of cultural exploration. On-site visitors may prefer quick facts, map-based navigation, or short summaries, whereas remote learners may favor narrative depth or thematic pathways. Device form factor also influences layout: mobile screens require vertical stacking and lightweight components, while desktop interfaces can support multi-panel layouts.

Despite adaptation, structural consistency is maintained to support stable mental models. Core behaviors (e.g., how cards expand, how related items appear) remain uniform across contexts to reduce learning effort.

This layer closes the pipeline from user goals to interface:

$$\mathrm{User} \mathrm{Task}\to \mathrm{Semantic} \mathrm{Components}\to \mathrm{UI} \mathrm{Components}\to \mathrm{Screen} \mathrm{Instance}.$$

3.4. Conceptual Diagram of the Framework

Figure 2 represents a hierarchical layered framework for modular information architecture in cultural heritage UX. The left branch illustrates the UI Components layer, including interface elements such as cards, previews, detail panels, map pins, and timeline elements, along with their progressive disclosure strategies and screen contexts (e.g., mobile on-site visit, desktop exploration, and thematic browsing). The right branch represents the Semantic Layer, beginning with user tasks—such as exploring artifacts, planning routes, or comparing experiences—and mapping them to corresponding content components (Artifact, Person, Place, Event, Theme). This structure visually emphasizes the separation and flow between user-driven semantic goals and their interface-level representations.

The diagram functions as a meta-model showing how user needs propagate through component definitions and UI patterns to concrete screens. It also serves as a validation tool: each user task must correspond to a clear semantic path and at least one viable screen configuration.

In summary, this layered modeling approach ensures cognitive efficiency, semantic clarity, and adaptability across user contexts. It provides a bridge between abstract user needs and practical interface structures, enabling coherent, scalable cultural heritage experiences.

4. Results

To assess how current cultural heritage platforms align with the proposed modular information architecture (IA), we compare TripAdvisor, Google Arts & Culture (GA&C), and Airbnb Experiences. We have clarified our rationale for selecting these three platforms in the previous section. Specifically, TripAdvisor represents the trip-planning phase of cultural heritage tourism, Google Arts and Culture represents the museum or curated content exploration phase, and Airbnb Experiences represents the active participation/experiential phase. By choosing these distinct archetypes (planning, exploration, and participation), we examine how modularity (or its absence) impacts user experience across the entire cultural heritage journey, rather than focusing only on a single type of platform. The goal is not interface critique but structural analysis of how each platform supports modularity, cognitive efficiency, and user-centered semantics.

4.1. Evaluation Criteria Based on the Proposed Framework

Drawing from our layered model, we evaluate platforms along five criteria:

• Content Modularity and Semantic Structure: presence of reusable, semantically defined units (e.g., Artifact, Place) and explicit relationships.
• Progressive Disclosure and Cognitive Load Management: use of summaries, staged detail, and low extraneous load [3].
• User-centered Navigation and Findability: alignment with mental models, intuitive labels, and multiple entry points [10].
• Personalization and Context Adaptation: tailoring of content based on user profile, context, or behaviour [28].
• Cultural Depth and Narrative Support: extent to which cultural meaning, interpretation, and narrative coherence are enabled.

These criteria operationalize our framework and allow structured comparison.

4.2. Platform Analysis and Comparative Summary

This section presents a qualitative analysis followed by a comparative summary table that synthesizes relative strengths and weaknesses across platforms.

TripAdvisor. TripAdvisor’s strength lies in planning-oriented navigation and social proof. Its content structure is largely flat: attractions are treated as atomic entities without semantic substructure or relationships. User-generated reviews dominate, offering subjective insight but limited interpretive depth. Progressive disclosure is present (e.g., expandable sections), though high review volume can increase cognitive load. Navigation aligns with tourism mental models (“Top places”, “Nearby”), but lacks heritage-specific pathways (e.g., themes, periods). Personalization is basic and context-light, mostly driven by location and recent searches.

Google Arts & Culture. GA&C most closely aligns with our framework. It employs a highly modular structure (Artists, Artifacts, Events, Collections) with clear semantic links, supported by knowledge-graph-level metadata. Progressive disclosure is strong: visual-first layouts and staged content reduce extraneous load while allowing depth when desired. Navigation supports multiple mental models (chronological, thematic, and geographic). Cultural depth is high due to curated narratives and multimedia storytelling. Personalization remains limited; most experiences are universal rather than adaptive, though mobile AR features add contextual relevance.

Airbnb Experiences. Airbnb offers structured, reusable UI-level modularity (consistent listing templates) but minimal semantic depth. Cultural meaning is embedded in host descriptions rather than structured components. Cognitive load is well-managed through clean layouts and concise summaries, prioritizing decision support. Navigation reflects activity- and logistics-oriented mental models rather than heritage exploration. Personalization and context adaptation (location awareness, behavioral recommendations) are strong but commercially motivated. Cultural interpretation varies by host and is not systematically supported.

To synthesize these observations,

Table 1. Comparative evaluation of the three platforms across the five design dimensions.

Dimension	TripAdvisor	Google Arts & Culture	Airbnb Experiences
Modularity	$★★$	$★★★★★$	$★★★$
Cognitive Load Mgmt.	$★★$	$★★★★★$	$★★★★$
Navigation & Findability	$★★★$	$★★★★★$	$★★★★$
Personalization	$★★$	$★★★$	$★★★★★$
Cultural Depth	$★★$	$★★★★★$	$★★$

provides a compact, comparative view of how TripAdvisor, Google Arts & Culture, and Airbnb Experiences align with the five core design dimensions of our framework. The star ratings are qualitative indicators derived from our comparative structural analysis of the platforms, representing each platform’s relative strengths and weaknesses across the five framework dimensions (Modularity, Cognitive Load Management, Navigation and Findability, Personalization, and Cultural Depth), and they are not intended as quantitative user metrics.

To further illustrate the broader relevance of our findings, we reference additional platforms such as Smartify (https://smartify.org/, accessed on 2 January 2026), a mobile museum guide app that exemplifies challenges in modularity and cognitive load through its linear content delivery, and Europeana (https://www.europeana.eu/, accessed on 3 January 2026), a cultural aggregator that, despite its rich metadata, often presents content in dense, non-adaptive formats. These examples reinforce the pervasiveness of cognitive overload and non-modular architectures across the digital cultural heritage landscape. This expanded set of examples supports the generalizability of our framework and highlights the need for modular, cognitively efficient design principles beyond the three primary platforms analyzed.

Overall, the comparison highlights the fragmented nature of current digital cultural heritage ecosystems. Each platform excels in different areas—GA&C in semantic depth, Airbnb in personalization, and Tripadvisor in social insight—yet none provides a fully modular, cognitively efficient, and user-adaptive architecture. This reinforces the need for integrated design approaches such as the framework proposed in this paper.

4.3. Insights and Gaps

The comparative assessment of TripAdvisor, Google Arts & Culture, and Airbnb Experiences surfaces a set of recurring structural patterns rather than isolated design quirks. While each platform demonstrates particular strengths when viewed through the lens of our framework, none of them implements a fully modular, cognitively efficient, and user-adaptive architecture for cultural heritage. In other words, the analysis does not simply differentiate “better” and “worse” designs but reveals complementary capabilities and persistent gaps that a next-generation cultural UX platform could strategically address. Three insights emerge:

• Divergent design philosophies: GA&C supports exploration and learning, TripAdvisor centers on planning, and Airbnb optimizes booking. This confirms that cultural UX requires balancing semantic depth with usability and decision efficiency.
• Need for semantic–social integration: User-generated content is rich but unstructured. Platforms lack mechanisms to integrate social insights into semantic models (e.g., thematic tags extracted via NLP), limiting hybrid discovery pathways.
• Opportunity for adaptive cultural UX: None of the platforms provide deep personalization grounded in cognitive load principles or user mental models. Modular IA could support adaptive pathways that adjust complexity, detail, or narrative style to user behavior and expertise.

Overall, the comparative analysis reveals a digital cultural heritage landscape marked by strong but uneven capabilities. TripAdvisor excels in usability and social proof but lacks the semantic depth and modularity required for rich cultural interpretation. Google Arts & Culture demonstrates the clearest alignment with our proposed framework, offering structured content units, progressive disclosure, and curatorial narratives—yet personalization and contextual adaptation remain underdeveloped. Airbnb Experiences, by contrast, provides highly polished interaction patterns and mature personalization, but the cultural layer is thin and dependent on host-generated descriptions rather than systematic semantic structures.

Taken together, these findings highlight a systemic gap: no existing platform integrates semantic modularity, cognitive load management, user-centered navigation, and adaptive personalization within a unified architectural model. Each platform contributes a piece of the puzzle—semantic richness, social context, or adaptive recommendations—but none synthesizes these into a cohesive, cognitively efficient cultural UX ecosystem.

This gap underscores both the relevance and the necessity of the proposed modular information architecture. It suggests that future cultural heritage systems can move beyond the current fragmentation by adopting component-based content models, cognitively aware interaction patterns, and explainable AI-driven adaptation. Such integration would not only enhance user comprehension and engagement but also enable scalable, cross-collection cultural experiences that support diverse cognitive styles and exploration behaviors.

5. Persona-Based Simulation of Cultural UX Journeys

To explore the practical implications of the proposed modular information architecture (IA), we conducted a structured simulation using five AI-generated personas representing diverse cognitive styles and cultural travel motivations. The simulation focused on a high-fidelity prototype (“CulturaPath”), allowing controlled evaluation of how modular content, progressive disclosure, and contextual adaptation support different user needs, as shown in Figure 3.

5.1. AI Personas

The personas capture distinct interaction patterns relevant to cultural heritage UX:

• Maria (Harmonious Explorer)—emotion- and narrative-driven; benefits from staged storytelling and soft guidance.
• Leo (Urban Decoder)—analytical and detail-oriented; requires rapid access to structured, factual content.
• Chloe (Aesthetic Curator)—visual-spatial; engages through imagery and clean layouts with minimal initial text.
• Nikos (Community Advocate)—ethically motivated; prioritizes contextual and social-impact information.
• Alex (Adventure Seeker)—action-oriented; responds best to concise logistics, map-based views, and difficulty indicators.

These personas collectively represent a spectrum of cognitive strategies (narrative, analytical, visual, ethical, and action-focused), enabling robust evaluation of the framework.

5.2. Simulation Setup

The simulation is design-oriented and exploratory rather than statistically conclusive; its goal is to probe the internal coherence of the IA and its alignment with diverse cognitive styles. Each persona journey was simulated across four stages: (1) Onboarding questionnaire; (2) Browsing of recommended experience cards; (3) Progressive expansion into detailed content; (4) A final conversion decision (e.g., saving or expressing interest).

We evaluated the interaction using:

• time-on-task per stage,
• estimated cognitive workload (NASA–TLX),
• simulated eye-tracking patterns based on established UX attention models,
• qualitative feedback synthesized per persona.

The simulation served to validate whether the IA structure effectively aligns with different mental models while minimizing extraneous cognitive load.

5.3. Results: Cross-Persona Patterns

Time-on-Task. Alex (action-driven) completed tasks the fastest, followed by Maria and Chloe, who engaged more with narrative or visual elements. Leo and Nikos showed the longest exploration durations due to deeper fact-based or ethical scrutiny.

Cognitive Load. NASA–TLX ratings remained low to moderate across all personas. Maria, Chloe, and Alex reported the lowest mental demand, reflecting effective progressive disclosure. Leo and Nikos showed higher effort, consistent with their preference for rich contextual or analytical content.

Attention Patterns. Simulated gaze behavior aligned strongly with persona traits: Maria focused on narrative blocks; Leo on headings and structured data; Chloe on imagery and layout; Nikos on community-impact modules; Alex on logistics, maps, and difficulty cues.

Qualitative Reactions. All personas reported low friction and high clarity. Leo requested faster bypass options to reach detailed content. Chloe emphasized aesthetic coherence. Nikos valued transparency and ethical context. Alex highlighted the usefulness of concise, actionable information.

5.4. Framework Validation

Across personas, the modular IA consistently:

• reduced extraneous cognitive load through progressive disclosure,
• aligned components with diverse mental models,
• enabled flexible interpretation without overwhelming the user,
• supported distinct decision-making pathways.

These findings indicate that a modular, component-based IA can accommodate a wide variety of cognitive strategies in cultural heritage contexts. The simulation provides preliminary evidence and an illustrative test of the framework’s adaptability and cognitive plausibility, but does not replace empirical validation with real users.

6. Conclusions and Summary of Contributions

This paper introduced a comprehensive theoretical framework for designing modular, user-centric information architectures in cultural heritage tourism. Addressing the challenge of diminishing attention and cognitive overload (Section 1), we grounded our approach in Cognitive Load Theory [3,15], mental models [9], and information architecture principles [10].

A key contribution is the integration of these theories into a design methodology: eliciting user tasks and mental models (Section 3), then translating them into a semantic content model (Section 3.1) that aligns with user cognition. This ensures the architecture is both logically structured and cognitively ergonomic.

We also proposed a three-layer architecture—Components, UI Components, and Screen Instances—governed by progressive disclosure and context adaptation (Section 3). This model connects atomic content units to consistent interface patterns and responsive screens, offering a reusable blueprint for heritage UX. By linking design decisions to cognitive theory (e.g., segmenting to reduce load), we provide theoretical justification for common UX practices.

The task-to-semantics mapping (Section 3.1) is a novel validation method, applicable beyond heritage to domains such as education or healthcare, ensuring content supports actual user goals.

Our comparative analysis (Section 4) of TripAdvisor, Google Arts and Culture, and Airbnb Experiences validated the framework’s relevance. GA&C’s semantic structure supports deep exploration, while TripAdvisor’s popularity highlights the value of social features and simplicity. Airbnb excels in usability but lacks depth. These findings underscore the need for a balanced approach—modularity, cognitive efficiency, and user relevance.

We also bridged HCI theory and heritage practice, adapting general UX principles to the interpretive nature of cultural content. Our component model is informed by standards like the CIDOC Conceptual Reference Model (CIDOC-CRM), a formal ontology for cultural heritage information that offers a practical tool for digital humanities projects.

Finally, the framework aligns with modern software paradigms, supporting implementation via component-based front-end architectures. In sum, our contributions include:

• A theory-driven framework for cultural UX;
• A validated modular content and interface model;
• Comparative insights that highlight design trade-offs and future opportunities.

7. Discussion and Future Directions

In Section 7, we now explicitly acknowledge the tension between AI-driven personalization and user data privacy. We note that such personalization often entails extensive user data harvesting, raising ethical and privacy concerns, and we emphasize that our modular architecture can mitigate this by supporting data minimization—clearly defining and limiting which user data components are utilized for recommendations—thereby enhancing transparency and user control over personal data. The proposed modular information architecture (IA) framework addresses a structural blind spot in many cultural heritage systems: content is often rich and heterogeneous, yet organized through ad hoc page templates or institution-centric taxonomies rather than user-centered semantic models. By grounding the architecture in cognitive load theory, mental models, and modular content principles, the framework offers a conceptually coherent way to reconcile informational richness with the reality of fragmented attention and mobile-first interaction.

From a cultural heritage perspective, the framework contributes to ongoing efforts to move beyond static collections interfaces and towards adaptive, experience-oriented systems. It complements existing work on digital storytelling, virtual museums, and AR/VR applications by focusing on the underlying content and interface structures that make such experiences cognitively accessible. In particular, the emphasis on components, UI components, and screen instances aligns well with contemporary component-based front-end development and with object-oriented approaches used in immersive media and AR experience design. Modular, metadata-rich content objects can be reused across channels (web, mobile, on-site kiosks, AR browsers), supporting cross-platform cultural journeys without redesigning the architecture from scratch.

At the same time, this work has clear limitations. First, it is primarily conceptual and analytical. The persona-based simulation demonstrates internal consistency and adaptability but does not substitute for large-scale empirical evaluation with real visitors or learners. Future studies should therefore test the framework through implemented prototypes in real museum or tourism contexts, measuring task success, engagement, and long-term recall. Second, the comparative analysis covers only three high-profile platforms and focuses on structural features visible at the interface and content level. Other systems (e.g., national aggregators, institutional repositories, niche cultural apps) may exhibit different trade-offs, and the internal architectures of commercial platforms are not fully observable. A further limitation lies in the use of AI-generated personas and simulated UX metrics. While grounded in established HCI models, such simulations necessarily simplify the diversity of real users and do not capture socio-cultural factors such as accessibility needs, language barriers, or varying levels of digital literacy. The framework should therefore be complemented by inclusive, participatory design processes that involve different audiences and stakeholders in co-creating navigation structures and component models.

Despite these constraints, the discussion highlights several practical implications. For designers and developers, the framework suggests prioritizing semantic component models and progressive disclosure before visual polish or feature additions. For cultural institutions, it underscores the value of investing in structured metadata and cross-collection ontologies (e.g., CIDOC-CRM, Dublin Core) as enablers of future adaptive experiences, including AR and location-aware services. For researchers, it opens a design space where cognitive theories, information architecture, and immersive media can be jointly explored through modular, empirically testable cultural UX prototypes [29].

From an implementation standpoint, the principles articulated in this paper are well-suited to modern component-based software architectures (e.g., React, Angular, Vue). These frameworks naturally support modular UI development, enabling reusable components, progressive disclosure, and contextual adaptation. For instance, an “Artifact” component can render summary or detail views depending on context and properties, mirroring our conceptual model. This aligns closely with Object-Oriented UX (OOUX), which emphasizes designing around user-recognized objects. A promising research direction is to empirically compare object-based designs with task-based or page-based approaches, measuring differences in comprehension, mental model formation, and cognitive load.

Human-Centered AI Implications

The modular information architecture proposed in this paper has direct implications for the design of Human-Centered Artificial Intelligence (HCAI) systems in cultural heritage, particularly for explainable and controllable recommender services. This aligns with broader HCAI agendas that emphasize reliable, safe, and trustworthy AI grounded in clear human goals and oversight [30,31]. Because content is modeled as explicit components (e.g., Artifact, Theme, Place, Experience) with well-defined attributes and relationships, the recommendation layer can operate on interpretable units rather than opaque page-level aggregates. This supports traceable decision-making: a system can explain a suggestion in terms of concrete components (“recommended because you engaged with similar themes and nearby places”) instead of inscrutable scoring functions.

In the CulturaPath use case, AI-driven personalization can leverage a range of signals while remaining cognitively and ethically legible. Behavioral signals include card impressions and clicks, dwell time on specific components, scroll depth, revisits, and completion of narrative- or logistics-focused flows. Contextual signals include on-site versus remote usage, time constraints, mobility preferences, and selected cognitive style from onboarding (e.g., narrative-driven Maria, logistics-driven Alex). Simple, interpretable models—such as weighted scoring, rule-based filters, or shallow decision trees—can be applied on top of the component graph to produce recommendations that are both actionable and explainable at the UI level (e.g., “Because you favor short, action-oriented routes, this experience is ranked higher”).

Modularity also facilitates controllability and human-in-the-loop oversight. At the user level, CulturaPath can expose direct controls that map cleanly to component attributes: sliders for preferred depth of information, toggles between narrative and factual views, filters for accessibility or crowding, and opt-out options for specific data uses. These mechanisms operationalize human-control principles articulated in HCAI frameworks and human–AI interaction guidelines, where users should be able to override system choices, correct recommendations, and understand the consequences of their feedback [32,33]. At the institutional level, curators and cultural professionals can review which components and relationships drive recommendations, pin or demote specific experiences, and audit how AI models interact with the underlying semantic model. This supports ongoing curator–AI co-evolution rather than fully automated curation.

From a regulatory perspective, modular IA can act as an enabling layer for compliance with emerging AI governance frameworks, including the EU AI Act. Even if most cultural heritage recommenders fall into lower-risk or limited-risk categories, their use of profiling, location data, or potentially affective signals places a premium on transparency, data minimization, and robust logging. By cleanly separating content components, user-profile attributes, and recommendation logic, the architecture makes it easier to document which data are processed, why particular items are suggested, and which human oversight mechanisms are in place. Explanatory UI components (e.g., “Why am I seeing this?” panels bound to specific component attributes) can be systematically designed rather than added ad hoc, aligning cognitive accessibility for users with traceability requirements for regulators.

In summary, the modular, component-based approach does not merely improve UX; it provides a structural foundation for human-centered, explainable, and controllable AI in cultural heritage. CulturaPath illustrates how AI-driven personalization can remain grounded in interpretable content objects and explicit user signals, supporting both richer cultural journeys and more accountable recommender behavior.

Looking ahead, personalization and adaptive interfaces remain a major avenue for future research. Cultural heritage platforms often underutilize personalization, especially in ways that are transparent and respectful of users’ cognitive constraints. Our modular design supports dynamic adaptation—for example, simplifying views for novices or enriching them for experts, or detecting behavioral indicators such as rapid scrolling and adjusting the interface accordingly. Future work should explore explicit, user-friendly adaptation cues (e.g., “Switch to expert mode?”) to maintain trust and agency.

Integrating user-generated content into structured systems is another opportunity. Platforms like TripAdvisor and Airbnb demonstrate the value of community input but lack semantic integration. Future systems could treat comments or anecdotes as structured components (e.g., UserComment), moderated and surfaced via relevance scoring. This raises questions about information quality, provenance, and trust, inviting research from CSCW and social computing.

Our framework also supports cross-institutional integration. By adopting metadata standards (e.g., Dublin Core, CIDOC-CRM), systems could aggregate content from museums, libraries, and archives. Challenges include schema alignment and semantic interoperability, but the potential for unified, scalable cultural platforms is significant. Intelligent search and filtering—for instance, querying “18th-century paintings in France” across multiple collections—would be essential to realize this potential.

Finally, empirical validation is critical. Prototypes should be tested with real users (e.g., museum visitors, students) to assess usability, comprehension, and engagement. Both qualitative feedback and quantitative metrics (e.g., time on task, retention, return visits) would inform iterative refinement. Longitudinal studies could assess sustained engagement and inclusivity—key goals for cultural institutions that seek not only to attract visitors but also to support meaningful, repeatable cultural journeys.

In conclusion, the proposed framework offers a foundation for building human-centered, cognitively efficient cultural heritage platforms. Future work should focus on implementation, testing, and refinement to ensure that technology enhances, rather than obstructs, meaningful exploration of cultural knowledge.