Mediating Perception and Participation: Abstract Urban Sculptures in Augmented Reality (AR) and Web3 Environments for Socially Sustainable Design

Abstract

Social sustainability in urban and architectural design depends on inclusive, participatory processes that empower communities to actively engage in shaping their environments. This study investigates how emerging digital platforms, specifically Augmented Reality (AR) and decentralized platforms built on blockchain technology (Web3), can function as instruments for broadening public participation and enhancing perceptual access to urban art proposals. An original algorithm generated nine digital abstract sculptures, each with descriptive attributes forming the basis for qualitative analysis across different visualization modes: traditional renderings, Augmented Reality environments, and NFT-based Web3 representations. Through participant voting, each digital sculpture accumulated a measurable level of preference that served to identify which sculpture was perceived as most successful within each visualization context. Comparative analysis revealed how distinct digital interactions shape perception, engagement, and inclusivity of feedback processes. Regression models further predicted voting outcomes, showing that different sculptural attributes played a dominant role depending on the type of visualization. Findings indicate that platform-specific technological affordances substantially shape participatory outcomes. Consequently, the study argues that careful analysis and selection of the digital platform must precede any participatory process, as platform-specific affordances fundamentally condition the inclusivity, accessibility, and overall effectiveness of public engagement in socially sustainable design.

1. Introduction

Social sustainability has become an increasingly central concern in contemporary urban and architectural design; it extends beyond the physical form of the built environment to encompass the social relations, inclusivity, and participatory structures that shape it [1]. Within this context, design processes are expected not only to respond to environmental and economic imperatives but also to empower communities to participate meaningfully in shaping their spatial surroundings. Participation and inclusion thus emerge as key dimensions of socially sustainable urban art development; they ensure that diverse voices contribute to the design and decision-making processes that define collective space [2,3,4].

In recent years, the expansion of digital technologies has opened new possibilities for public engagement in art and design. Digital visualization tools, interactive media, and online platforms have begun to transform the way urban and architectural ideas are communicated, negotiated, and experienced. Among these emerging tools, Augmented Reality [5,6,7] and Non-Fungible Token (NFT), as a specific type of digital asset built on top of Web3 [8,9], present particularly interesting opportunities for rethinking participation. Augmented Reality enables users to perceive and interact with virtual elements embedded within real environments; NFTs introduce new modes of digital ownership, distribution, and authorship that may redefine how design proposals circulate and gain public visibility. When applied to participatory design, these technologies have the potential to broaden access, diversify engagement, and foster new forms of collective authorship, particularly as ongoing technological advancements continuously reshape the behavioral patterns of the audiences with whom we engage. Therefore, it is essential to identify new approaches that foster the creation of innovative communication tools and channels capable of improving both the level and quality of participation [10].

This research investigates how Augmented Reality and NFT Web3-based platforms can serve as experimental instruments for expanding participatory practices in urban digital art design. Specifically, it examines how these digital modes of representation influence public perception, engagement, and the inclusivity of feedback mechanisms. The main goal of the research was to determine whether different digital platforms for participation lead to different participatory results; in other words, whether the choice of platform itself affects the nature of public engagement and decision-making outcomes. The central hypothesis posited that each platform attracts different segments of society, i.e., groups characterized by varying degrees of digital literacy and familiarity with technology, which in turn shape distinct evaluative criteria that participants consider important when forming their decisions.

The whole research has been conceived as a second round of an architectural or art competition for a defined urban fragment, in which the author(s) have already been selected, yet present several variations in the proposed design for the public to evaluate and decide upon. The study is conceived as a replication of a real architectural or art competition. In practice, competitions function by having authors submit their work, which is then evaluated either by a professional jury or, ideally, by the broader public. In this research, two parallel architectural competitions were simulated: one in which the public voted on projects presented through renderings, and another in which voting was based on Augmented Reality presentations. The aim was to demonstrate that the presentation medium affects both the level of public engagement and the final selection outcome—both of which were confirmed by the study.

In this context, the NFT model served as a control group, functioning as a neutral format of digital exchange that does not directly influence the perceptual aspect of decision-making. Additionally, due to the nature of registration on the Objkt platform, the NFT environment ensures a high level of user anonymity, while simultaneously requiring a higher degree of technical proficiency. For this reason, demographic data such as the gender and age of NFT participants could not be obtained; however, it can be stated with a reasonable degree of certainty that their level of technical literacy is above average.

In the context of architectural competitions, particularly in the second evaluation phase, a selection of nine proposals is not unusually small. On the contrary, it is common for the number of shortlisted entries at this stage to be comparable or even smaller, depending on the structure and scope of the competition.

To foster this idea of an informed public choice within a controlled creative framework, all presented digital sculptures were generated using the same algorithmic procedure. Furthermore, the entire process of postproduction, encompassing rendering, Augmented Reality integration, and NFT tokenization was conducted by the same team member to ensure a consistent aesthetic and technical approach across all platforms.

To explore objectives stated above, first, an algorithmic process for generating minimal networks and surface envelopes, based on a pseudo-Steiner algorithm [11], was employed to create nine abstract sculptural forms designed for potential integration into public urban contexts. For each generated sculpture form, a set of descriptive attributes was produced, including identifiers, basic geometric parameters (such as the number of points, faces, areas), as well as statistical indicators derived from these values (distance and elevation dispersions, and object complexity). These combined parameters formed the basis for comparing the digital sculptures and provided input for both quantitative and qualitative analyses.

These sculptural forms were then presented to three distinct participant groups, each experiencing one of three types of visual representations: traditional two-dimensional (2D) renderings, Augmented Reality environments, and NFT-based digital visualizations within the Web3 environment. Participants were invited to evaluate the artworks through a simple voting procedure; each sculpture accumulated a total number of votes, providing a comparative basis for analyzing how different visualization technologies affect participants’ perceptions and decision-making.

Building upon this context, we hypothesize that the type of visual representation significantly shapes the character and depth of public engagement in participatory design processes. Specifically, we expect that each visualization modality—static renderings, augmented reality environments, and NFT-based digital formats—activates different perceptual, cognitive, and interpretive pathways for participants, thereby influencing how they evaluate design attributes and how they understand their own role within the decision-making process. We further posit that technological affordances embedded in each platform are not neutral but instead conditional in terms of varying degrees of accessibility, inclusivity, and agency. In parallel, we hypothesize that predictive modelling of engagement patterns will reveal distinct relationships between sculptural attributes and participant responses depending on the platform through which the work is experienced. Finally, given the absence of prior research integrating NFT-based visualization into participatory assessment frameworks, we anticipate that this study will open new methodological directions and invite reconsideration of how emerging digital environments can expand the theoretical and practical landscape of participatory design.

2. Theoretical Background

The growing influence of technology on contemporary modes of perception, creation, and communication has profoundly reshaped the fields of art, design, and architecture. As Rinehart & Ippolito (2022) observed, technology represents one of the central challenges to twenty-first-century creativity, particularly because much of new media art depends on transient software and hardware systems that constantly evolve and become obsolete [12]. This dependency introduces not only technical but also conceptual issues related to perception, participation, preservation, authorship, and cultural memory. The digital turn in creative practice has expanded the conceptual and methodological boundaries of design, enabling algorithmic and generative processes that incorporate a variety of spatial and contextual parameters. These developments have led to new forms of artistic and architectural expression in which computational logic, data, and interactivity serve as integral design components [13]. At the same time, digital visualization tools and interactive online platforms, particularly Augmented Reality and Web3 technologies, have reshaped the ways in which urban design ideas are communicated and collaboratively developed, especially within participatory processes related to urban environments.

2.1. Participation and Art in Public Spaces

Citizen participation as a mechanism for improving public spaces has a long-standing tradition, although its implementation has often remained at an experimental level [14]. Over time, participation has become a central concept in discussions on urban governance, sustainability, and spatial equity [2]. However, the practical realization of participatory models frequently encounters significant challenges. The most common obstacles relate to the accessibility and dissemination of participatory tools and questionnaires, as participatory processes often involve a limited circle of directly interested stakeholders, rather than the broader public [15,16]. Consequently, the outcomes of such processes can reflect the views of a narrow group rather than the diversity of urban society. It is also important to recognize that participatory methods differ considerably in the stage and extent of public involvement. Some frameworks advocate continuous public inclusion throughout the process, from problem identification to solution development, while others involve citizens only at the stage of evaluating pre-defined results. These variations are frequently associated with motivational barriers, limited access to digital or physical participatory platforms, and financial constraints that restrict the scope of engagement [17,18]. Such factors indicate that participation should not be observed solely as a procedural tool but as a complex social practice that requires both institutional support and a well-designed methodological framework. Beyond the issues of accessibility and citizen motivation, another critical aspect concerns the format of participation itself—specifically, how to design participatory formats that are both inclusive and contextually relevant. The optimization of participatory tools is essential for ensuring that diverse segments of the population are meaningfully represented. Despite these difficulties, contemporary research provides numerous examples of successful participatory initiatives that have contributed to urban sustainability. These initiatives encompass various fields relevant to the shaping of urban environments, including architecture [2,19], infrastructure [20], the redevelopment of historical heritage [21], and urban renewal [22,23]. Collectively, these studies demonstrate that participation, when properly structured, can strengthen social cohesion, improve spatial functionality, and enhance the aesthetic and symbolic dimensions of urban life.

2.2. Render, Augmented Reality and Web3: Visual Representation of Digital Art

Digital art projects often rely on audience engagement and interpretation, frequently concealing their full meaning at first glance [24]. This makes the mode of visual representation a critical factor in shaping how such works are perceived and experienced. The most conventional method of visualizing digital 3D models is rendering, which produces a static two-dimensional image of a three-dimensional object viewed from a predefined angle. While this technique allows for precise control of lighting, texture, and composition, it also confines perception to a fixed viewpoint, emphasizing aesthetic over experiential qualities [25]. Consequently, the physical context in which a rendered image is displayed continues to play a decisive role in how the digital artwork is interpreted.

The rapid evolution of mobile smart devices and immersive media technologies over the past two decades has, however, transformed the possibilities for visualizing, exhibiting, and interacting with art [26]. These developments enable experiences that extend beyond the physical realm into virtual or hybrid spaces, fundamentally reshaping the relationship between artwork, viewer, and environment. Augmented Reality, in particular, situates digital content directly within real-world settings, allowing users to perceive and interact with virtual objects as though they were physically present. In this context, both the virtual model and the surrounding environment become integral to the artistic experience. Owing to these properties, Augmented Reality has been widely adopted in architectural, urban, and engineering visualization [27,28,29,30,31,32], in museum-based engagement [33], and for visualizing cultural heritage and digital art in situ or remotely [34].

In contrast, Internet-based digital art operates within a fundamentally different spatial logic. While it can be viewed from anywhere and at any time, independent of physical location [24], the spatial context can still influence interpretation by providing cultural or thematic framing. The rise of Web3 technologies, most notably Non-Fungible Tokens (NFTs), has brought a new dimension to this online environment. Since 2021, NFTs have propelled digital art into mainstream attention, as digital images and collectibles achieved record-breaking auction prices and inspired a proliferation of dedicated marketplaces [24]. NFTs are not artworks per se but rather blockchain-based records of ownership that authenticate a specific digital asset, thereby rendering it non-fungible—unique and irreplaceable [35].

The integration of Web3 technologies into art and design has introduced decentralized mechanisms for authorship, ownership, and distribution. This paradigm shift aligns closely with participatory and socially sustainable design values, offering novel ways for audiences to engage, co-create, and redistribute digital content. Within the emerging Web3 environment, such interactions are further enhanced through the convergence of immersive multimedia computing, artificial intelligence, ubiquitous networking, and digital identity [36]. These systems collectively foster a participatory, content-rich ecosystem in which digital artworks can circulate beyond traditional institutional or geographic boundaries.

Recent research indicates that Web3 principles play a central role in NFT art creation, with blockchain features, such as transparency, openness, and authentication, shaping how artworks are owned, traded, and experienced [37]. The characteristics of blockchain technology have enabled NFTs to merge with art, offering new perspectives and playing an important role in the digital art economy [38]. Within digital culture, NFTs have also led to the emergence of a unique aesthetic sensibility that can be understood as an amalgamation of artistic and technological elements [39]. This is largely because NFTs, defined as blockchain-based tokenized digital assets, have (re)shaped the ways users perceive and interact with digital collectibles, including physical art, music, and 3D assets [40,41]. The users who participate in specific NFT collections may form communities whose engagement is driven not only by fundamental technical aspects such as ownership rights, but also by elements of exclusivity, including social status and access to specific communities and events [40].

These characteristics underscore the potential of Web3 platforms to enhance participatory practices, fostering inclusive engagement, collective authorship, and greater transparency in digital art experiences [37].

2.3. Research Gap and Objectives

Within the scope of this research, particular attention is given to public art as an increasingly significant element of urban spaces [42]. The study focuses on assessing the adequacy of visual formats for presenting digital abstract sculptures in public environments, exploring how the form of digital artistic visualization influences public perception and engagement. Public art is often regarded as an essential medium for expressing collective identity and shaping the atmosphere of shared spaces. Existing literature has documented multiple benefits of art in public contexts, including the enhancement of spatial ambiance, the transmission of symbolic meaning, and the promotion of psychological well-being and social connectedness among citizens [43,44]. Moreover, recent studies underline that artistic installations can stimulate social awareness and civic responsibility by addressing pressing societal and environmental issues. Through this role, public artworks evolve beyond aesthetic objects into catalysts of collective consciousness and instruments of civic engagement [45,46,47].

Although architectural and artistic design competitions have been extensively discussed in academic literature [48,49,50,51], the influence of representational formats on competition outcomes has received considerably less attention, particularly in contexts involving broader public participation. In most studies, visualization is regarded as a neutral medium for communicating design intent rather than as an active variable that can shape aesthetic judgment and decision-making.

This study addresses that gap by demonstrating that the choice of representational platform, whether static rendering, NFT-based animated visualization, or Augmented Reality spatial representation, directly affects both the composition of participating audiences and the nature of their evaluative responses. In this context, the present study focuses on evaluating the adequacy of visualization formats used to represent abstract digital sculptures during participatory assessment processes.

The findings are expected to contribute to the broader discourse on integrating digital visualization technologies into participatory urban design processes, demonstrating their potential to bridge the gap between complex artistic concepts and public understanding. Moreover, by analyzing how different formats mediate the relationship between digital art and its prospective spatial context, the study aims to propose guidelines for improving the accessibility and interpretability of digital artworks within urban environments.

The research is based on the evaluation of algorithmically generated abstract digital sculptures, which were publicly presented through three visualization modalities: traditional rendering, Augmented Reality, and Web3 (NFT) platforms. Each of these environments offered distinct perceptual and participatory conditions, enabling the comparative analysis of user engagement, preference formation, and the social dynamics of digital mediation. The Web3 sculptures were hosted on the Objkt platform, an NFT marketplace operating on the Tezos blockchain, chosen for its accessibility to artists and its minimal transaction (“gas”) fees. This ecosystem, which promotes the idea of “art for everyone,” provided an experimental framework for exploring how decentralized digital platforms can facilitate broader audience participation and new forms of authorship in the evolving field of urban digital art design.

3. Materials and Methods

The study was conducted in four distinct phases (Figure 1). Phase I involved the creation of a series of digital sculptural forms developed through a custom parametric algorithmic workflow (see Section 3.1). In order to generate the abstract sculptural forms used in this research, a custom parametric workflow was developed. One of the central components of this workflow is a pseudo-Steiner algorithm [11] for generating minimal networks and their surface envelopes, described in detail below. This phase established the formal and generative foundations of the study, ensuring that all subsequent visualizations were derived from a consistent computational logic.

Phase II comprised the postproduction process, during which three different visualization modalities were produced: classical rendering, Augmented Reality, and NFT-based videos within the Web3 environment (see Section 3.2). These modes were intentionally selected to represent distinct ontological and perceptual conditions of digital representation, ranging from static simulation to immersive spatial augmentation and tokenized digital mediation.

Each digital sculpture was thus presented through the three aforementioned types of visual representations, each constructing a different relational framework between the object, the viewer, and its surrounding context (Figure 2). Augmented reality provided an environmentally contingent visual experience, situating the sculpture within the user’s immediate physical surroundings and creating a hybrid condition of presence that merged the tangible and the virtual. This dynamic interplay of spatial, temporal, and perceptual layers reflects what has been termed digital phenomenology, the experiential negotiation between material space and its digitally augmented counterpart. In contrast, the NFT-based video representation, while time-based and potentially immersive, remained a pre-rendered, fixed digital artifact. It foregrounded the notion of media materiality, emphasizing the autonomy of the digital object as both artwork and data token, an entity existing within the symbolic and economic infrastructure of the blockchain rather than in physical space. The classical rendering, as the most static of the three modalities, offered a formally resolved yet context independent visualization. It reinforced the conventions of architectural representation by emphasizing materiality, composition, and perspectival control within a simulated environment. However, in doing so, it also exposed the limits of traditional visualization methods in capturing the participatory and experiential dimensions of design.

Phase III encompassed the conduction of surveys corresponding to each visualization mode, administered to distinct social groups in order to examine evaluative responses. This phase sought to determine whether the choice of platform, understood here as both a technological and social interface, directly influences participatory outcomes. Finally, in Phase IV, a series of regression models was developed to analyze and predict the voting percentages associated with the Render, Augmented Reality, and NFT categories. Three groups of regression models were constructed, each targeting one of the three output variables. The input parameters in the analysis consisted of the sculptures’ attribute data, while the output variables represented the percentages of votes received for NFT, Render, and Augmented Reality visualizations.

3.1. Algorithmic Generation of Minimal Networks and Surface Envelopes Using a Pseudo-Steiner Algorithm

The generation process in this research is inspired by the built project Collab Tape by the art collective Numen/For Use, executed in a park setting in Novi Sad (Serbia’s second-largest city, with an estimated 370,000 residents based on the 2022 census), where existing trees served as anchor points for a spatial membrane [52]. In the computational model, an initial set of points is randomly distributed, representing future supports of the structure. These points are processed using the nearest-neighbor principle, forming clusters that act as potential anchors for the minimal surface (Figure 3a).

The overall form of the structure is generated using a custom pseudo-Steiner algorithm [11], which approximates the behavior of Steiner trees through iterative geometric operations. Specifically, the algorithm links each active point to its closest neighbor, introducing intermediate points between them to minimize connection lengths (Figure 3b). This approach draws on established heuristics developed for Steiner networks, where the closest terminals are progressively connected and new points—often placed at midpoints—are added to reduce overall path length [11,53]. The resulting network serves as a structural skeleton that is then relaxed through spatial optimization, further minimizing total length and producing junctions typical of minimal networks (Figure 3c).

Once the relaxed skeleton is obtained, a continuous surface envelope is generated around it, resulting in a smooth minimal-surface-like geometry that connects all anchor points and produces a coherent spatial structure (Figure 4). This workflow aligns with principles of generative design applied to minimal surfaces in parametric modeling, as demonstrated by Gökmen (2022), who explored algorithmic approaches to develop minimal-surface-based tectonics [54].

For each generated segment and node, a set of descriptive attributes is produced, including identifiers, basic geometric parameters (such as the number of segments, points, and faces, as well as areas and total lengths), metric values (distances and elevations), and statistical indicators derived from these values (means, dispersions, and coefficients of variation). These combined parameters form the basis for comparing the generated configurations and provide input for both quantitative and qualitative analyses, which are presented and discussed in Section 4.

For analytical clarity, the following parameters were defined to describe the geometric and statistical properties of each configuration. Together, they quantify the internal organization and structural variability of the generated minimal networks and their surface envelopes. These parameters serve exclusively as evaluation criteria for assessing the generated structures, providing a consistent analytical basis for comparing their geometric characteristics and for interpreting the results discussed in the subsequent sections:

• Total Points

Total Points represent the total number of characteristic points that define the shape. A higher number indicates a more detailed and geometrically refined form. In architectural modeling, this parameter corresponds to the resolution or granularity of the generated structure.

$$TotalPoints=N_p$$

where $N_p$ is the total number of points (coordinates) defining the geometry.

• Area

Area represents the overall surface of the analyzed object. Larger values indicate a broader spatial coverage or a more extensive geometric footprint. This parameter is one of the most direct indicators of the object’s size and proportional dominance.

$$Area=\sum _{i=1}^n A_i$$

where $A_i$ is the area of each polygon, and $n$ is the total number of polygons.

• Distance Dispersion

Distance Dispersion measures how evenly points are distributed around the center of mass. A value of zero indicates a perfect circle (in 2D) or a perfect sphere (in 3D), meaning all points are equidistant from the centroid. Higher values show increasing irregularity and morphological complexity of the shape.

$$DistDisp=\sqrt{{\displaystyle \frac{{\sum }_{i=1}^n {\left(d_i-\underset{\_}{d}\right)}^2}{n}}}$$

where $d_i$ is the distance of each point from the centroid, and $\underset{\_}{d}$ is the mean distance.

• Complexity (Projected Intersections Index)

The Complexity parameter quantifies the degree of structural or visual intricacy in the projected geometry. It measures how many times the projected edges of a form intersect, normalized to the most complex case in the dataset. By adding 1 to both numerator and denominator, this modified formula ensures that even the simplest shapes (those with no intersections) receive a baseline value greater than zero, making the parameter suitable for comparative morphological analysis.

$$Complexity={\displaystyle \frac{I_p+1}{I_{max}+1}}$$

where $I_p$ represents the number of projected edge intersections for the given geometry, and $I_{max}$ is the maximum number of intersections observed within the dataset.

• Elevation Dispersion

Elevation Dispersion describes how much the form varies vertically. Low values correspond to flat or uniform surfaces, while higher values indicate stronger relief articulation, slopes, or topographic dynamics. It captures the degree of vertical differentiation within the geometry.

$$ElevDisp=\sqrt{{\displaystyle \frac{{\sum }_{i=1}^n {\left(z_i-\underset{\_}{z}\right)}^2}{n}}}$$

where $z_i$ represents the height of each point, and $\underset{\_}{z}$ is the mean height.

To enable a comparative graphical representation of the attributes for each sculpture, the attribute values were normalized to a 0–1 scale (Figure 5).

3.2. Types of Visual Representation

Each of the nine parametric sculptures was visualized through three complementary modes of representation: (1) static renders, (2) Augmented Reality models, and (3) 360° videos presented in a Web3 environment. To ensure visual consistency and contextual relevance across all media, a single real-world background was used, representing a public urban park adjacent to the university campus. To create this background environment, the site was captured using the Insta360° digital camera to generate high-resolution HDRI (High Dynamic Range Imaging) maps. The park was photographed during the blue hour to achieve a subtle nocturnal atmosphere that enhances the perceptual prominence of the sculptures against the ambient lighting. These HDRI maps were subsequently used both for lighting and environmental reflections in the rendering and compositing stages.

For the traditional renders (Figure 6), each sculpture was positioned within the same HDRI-based scene and rendered as a static two-dimensional visualization. For the augmented reality representation (Figure 7a), the 3D models were exported in .usdz format to enable viewing on portable devices, allowing users to explore and position the sculptures at a 1:1 scale within real-world urban environments. The Augmented Reality deployment aimed to evaluate user responsiveness and perceptual interaction with abstract parametric forms integrated into the actual urban setting.

Finally, for the Web3 presentation, interactive 360° videos were generated to provide immersive access to the sculptures in a virtualized public park environment. The animations were rendered in DaVinci Resolve using Fusion compositing to integrate 3D models with HDRI panoramas. Camera paths were scripted to perform a full 360° orbit around each sculpture, with smooth easing and depth-of-field effects to simulate cinematic motion. The final outputs were encoded in H.265 format for optimized streaming performance in Web3-based virtual galleries.

The collection of digital art sculptures, titled Abstract Urban Sculptures was created and published on the Objkt platform, one of the NFT marketplaces (Figure 7b). The collection featured nine digital sculptures, each accompanied by metadata describing its distinctive attributes.

3.3. Survey Design

To identify preferences regarding specific modes of visual representation—renders, Augmented Reality, and NFTs—the research employed a series of online surveys. Three independent studies were conducted to ensure that participant groups did not overlap, thereby avoiding potential bias and guaranteeing that each group evaluated only one type of representation. The surveys were conducted over the 7-day period, from 25 September 2025 to 1 October 2025, aiming to collect quantitative data on participants’ preferences and perceptions. This timeframe allowed for sufficient data gathering while ensuring consistency across the responses.

The studies focusing on renders and Augmented Reality were carried out using Google Forms questionnaires, which were distributed to participants via email or messaging systems (70 invitations each, with possibility for each invitation to be further distributed by participants). Both questionnaires included several general questions concerning gender, age, and the level of technological literacy. At the beginning of each survey, participants were required to provide informed consent, which clearly explained the purpose of the research and the intended use of the collected data, emphasizing that no personal information would be disclosed or used for any purpose beyond the study. After completing the introductory section, participants proceeded to the part of the questionnaire designed to assess user preferences for abstract sculpture designs visualized either as rendered images or through Augmented Reality, depending on the assigned group. Each participant was presented with the following scenario:

“This research is based on a simulation of a competitive process for the realization of abstract sculptures intended for public space. It is assumed that a single artist has reached the final stage, at which point it becomes the responsibility of the public, or other relevant stakeholders, to evaluate the individual sculptures created by the artist.”

Participants were then asked to evaluate each sculpture by selecting YES—indicating “I would like this sculpture to be installed in public space”—or NO—indicating “I would not like this sculpture to be installed in public space. “They were allowed to provide multiple YES or NO responses; in other words, they were not limited to selecting only one sculpture. For the group assessing render-based representations, visualizations were embedded directly in the questionnaire. The Augmented Reality group was provided with links to download .usdz files, enabling them to view each sculpture in augmented reality using their own devices. As a result, the group of respondents who completed the survey evaluating the sculptures presented in render format comprised 63 participants, while the group that assessed the sculptures in Augmented Reality included 33 participants.

The third group of participants was assigned to evaluate the sculptures represented as NFTs. This stage of the study required a slightly more elaborate methodological framework, given the specificity of the medium and the technical requirements involved in NFT presentation and access. The Web3 survey was conducted on the Objkt platform, one of the NFT marketplaces operating within the Tezos blockchain. This environment was chosen because Tezos has established itself as a blockchain ecosystem that attracts artists and digital creators, while also being known for its exceptionally low transaction costs, or gas fees, which are almost negligible.

Due to the very nature of blockchain technology and the anonymity inherent to decentralized platforms within Web3 ecosystems, it was not possible to obtain demographic data such as the gender or age of NFT participants.

The process of voting and collecting NFTs required several conditions to be met. To access Objkt as the platform where the collection was hosted, participants first needed to possess a functional crypto wallet compatible with the Tezos blockchain. This wallet served as a means of user identification through a unique public key hash, which had to be recognized by the Objkt platform when a signing transaction was initiated on the blockchain.

In order to participate in voting, users were required to collect a specific NFT. This process involved a blockchain transaction through which the NFT was minted (created as a digital asset on the blockchain) and automatically transferred to the participant’s wallet.

Given the technical complexity of this process, it can be stated with a high degree of certainty that the participants possessed above-average levels of technical literacy.

Furthermore, since the act of selecting a sculpture in NFT form required its minting, meaning the initiation of a blockchain transaction that creates the desired token and places it into the user’s digital wallet, the low cost of this process in Tezos cryptocurrency (ꜩ) was a key factor in ensuring accessibility. The sculptures themselves, as non-fungible tokens, had no assigned cryptocurrency value since their price was set at 0 ꜩ. As a result, the total cost of participation was limited to the minimal gas fees necessary to execute the transaction.

To introduce the survey to participants, a collection titled Abstract Urban Sculptures was created on the Objkt platform. The collection featured nine digital sculptures, each accompanied by metadata describing its distinctive attributes. Participation in the survey was enabled through a smart contract created with an open edition system, allowing an unlimited number of non-fungible tokens to be minted during a defined period. Once this period ended, the final number of minted tokens reflected the level of interest generated among participants. This model, common within the NFT context, differs from the fixed edition approach in which the number of available tokens is predetermined, thereby defining their maximum supply in advance.

During the seven-day survey period, each unique Tezos wallet could collect—effectively cast a vote for, up to nine different sculptures per transaction. Identical non-fungible tokens could not be collected multiple times within a single transaction. To reach a broader audience, additional promotion took place within the Web3 community gathered on the X platform. A total of 300 direct invitations were sent to individuals who regularly engage in the ecosystem through NFT creation and collecting. The invitation contained essential information about the experiment, a link to the Objkt collection, and a request for recipients to share the project within their own X accounts. This approach generated additional visibility. By the conclusion of the survey, the official post about the experiment had reached over 14,000 people and had been reposted more than 120 times. As a result, 208 individuals participated in the Web3 survey, collectively minting 752 non-fungible tokens. Section 3.4 presents the results obtained from all three conducted surveys.

A notable observation from the study is the lower response rate for the Augmented Reality-based survey, which included 33 participants out of 70 direct invitations, compared to the render survey, where 70 direct invitations led to 63 participants, and the Web3 survey, which included 208 participants from 300 invitations. This outcome can be attributed to the higher technical complexity and requirements of engaging with Augmented Reality presentations, which demand compatible devices, sufficient storage, and user familiarity with augmented environments. The lower Augmented Reality response rate validates our observation that technical prerequisites (specifically the requirement for iOS devices and storage space) significantly restrict inclusivity. Conversely, the high volume of NFT responses highlights the “Art for everyone” ethos and high engagement levels of the Web3 community.

The Augmented Reality sample is considered relevant and sufficient because the invited population was drawn from the same pool of colleagues and students as the render survey, yet the response remained lower due to these technological barriers. Obtaining large samples for specialized technologies like Augmented Reality and blockchain-based voting is inherently challenging due to the specific “digital literacy” required.

All participants were selected as relevant stakeholders (including students and professionals from various fields) providing diverse and informed perspectives. Moreover, the study’s design emphasizes comparative patterns across different platforms rather than general population statistics. This finding represents a key early result of the study, highlighting that the complexity of a digital platform can significantly influence participation levels and, consequently, the inclusivity of public engagement processes. The research was designed as a simulation of a competitive process to identify trends in preference and engagement behavior, rather than to establish demographic representativeness.

3.4. Regression Model Development

A series of regression models were developed to analyze and predict the voting percentages associated with the Render, Augmented Reality and NFT categories. Three groups of regression models were constructed, each targeting one of the three output variables. The input parameters used in the analysis are presented in

Table 1. Number of positive (yes) votes within each survey.

ID	“Yes” Votes Render	“Yes” Votes Augmented Reality	“Yes” Votes NFT
0	33	32	147
1	29	11	79
5	35	11	92
6	38	26	74
7	29	17	72
13	35	10	70
20	45	19	71
23	34	15	70
30	21	23	77

, while the output variables represent the percentage of votes received by NFT, Model, and Augmented Reality. To enable consistent numerical processing and comparison across models, all percentage values were normalized to the range [0, 1].

All numerical experiments were conducted using the Python version 3.11.7 programming environment, with the PyCaret version 3.3.2 [55] framework employed as the primary tool for automated machine learning (AutoML). PyCaret is used to automatically construct multiple data processing pipelines incorporating feature scaling, model training, and performance evaluation.

To address potential multicollinearity and improve computational efficiency, Principal Component Analysis (PCA) [56] was applied to the input dataset. Components explaining 95% of the total variance were retained, ensuring adequate representation of the original features while reducing the dimensionality of the data.

Finally, feature relevance and model explainability were examined using the SHAP (SHapley Additive exPlanations) framework [57,58,59]. This approach quantifies the relative contribution of each input variable to the model output, providing a transparent foundation for subsequent result interpretation and discussion.

4. Results

The first part of Section 4 presents the numerical outcomes obtained directly from the three conducted surveys. While these quantitative indicators provide valuable information, they do not offer direct answers regarding the validity of specific visualization types in relation to the given scenario. Consequently, the second part of Section 4 focuses on the presentation of predictive models, which are designed to assess and define the appropriateness of different visualization methods in the context of the scenarios examined within this study. By integrating both direct survey data and model-based predictions, this approach allows for a more comprehensive understanding of visualization effectiveness, bridging the gap between empirical observations and theoretical evaluation.

4.1. Survey Results

The presented results are based on three independent surveys, each examining a different type of visualization (Appendix A). These surveys involved varying numbers of respondents, reflecting differences in participation across the studies. Public engagement within the surveys for different types of visualizations (Render, Augmented Reality, and NFT) amounted to 63, 33, and 208 participants, respectively. The varying number of respondents suggests that different modes of presentation influence distinct levels of public willingness and engagement—an issue further elaborated in Section 5.

As the number of participants differs across surveys, the absolute counts of affirmative (“yes”) responses also vary (Table 1). It is important to emphasize that respondents were allowed to select multiple sculptures with a “yes” response; consequently, the values reported in Table 1 represent the cumulative total of all affirmative responses.

For easier visual interpretation, Figure 8 presents the percentages of affirmative responses for each of the three surveys. This graphical representation allows for a clear comparison of participants’ preferences across visualization types, despite the differences in sample size.

4.2. Feature Analysis Based on Regression Model

Based on the previously defined dataset and variable structure, a series of regression models were developed to analyze and predict the voting percentages associated with the Render, Augmented Reality, and NFT categories.

Due to the relatively small number of data points, unknown statistical distributions of the input data, we decided against using rigorous hypothesis testing. Instead, we opted for training a plethora of prediction models in each particular case, comparing them against a predefined set of performance measures, and choosing the most appropriate one. We then proceed to analyze the contribution of different input features using Shapley value analysis [60]. This procedure is in accordance with the primary aims and objectives of our work, where rigorously derived prediction models are not a goal in their own right, but served as an auxiliary analytical tool to formally identify dominant features and provide a mathematical basis for their architectural and artistic interpretation.

The regressor models were trained using PyCaret AutoML library. The following models were derived in each particular case: Extra Trees Regressor [61], Light Gradient Boosting Machine (LightGBM) [62], Dummy Regressor [63], Random Forest Regressor [64], Gradient Boosting Regressor [65], Decision Tree Regressor [66], AdaBoost Regressor [67], Least Angle Regression (LARS) [68], Linear Regression [69], Ridge Regression [70], Bayesian Ridge Regression [71], Lasso Regression [72], LassoLARS [68,72], Elastic Net [73], Orthogonal Matching Pursuit [74], Passive Aggressive Regressor [75], Huber Regressor [76]. The trained models were compared against one another using Mean Absolute Error (MAE), Mean Absolute Square Error (MASE), and R2 measure.

4.2.1. Render Vote Prediction Model

For the analysis of Render vote predictions, the Extra Trees Regressor [61] was identified as the best-performing model. Its evaluation metrics indicate a high level of predictive accuracy, with a low mean absolute error (0.0134) and mean squared error (0.0003). The root mean squared error (0.0157) further confirms the model’s stability, while the median absolute error (0.0141) shows consistent performance across observations. Overall, the results suggest that the model provides reliable predictions of Render votes with minimal deviation.

The results suggest that Total Points is the primary driver in predicting Render votes. According to the feature importance plot (Figure 9), Total Points contributes most to the model’s predictive accuracy, followed by Area, which shows a moderate positive effect. The SHAP summary plot (Figure 10) further confirms this relationship, indicating that higher values of Total Points strongly increase predicted Render votes. In contrast, Complexity and Elevation Dispersion exhibit limited influence on the model output, suggesting their contribution to vote prediction is marginal. The predicted versus actual values plot demonstrates that the model approximates Render votes with high accuracy, although slight overestimation occurs at lower vote levels and minor underestimation at the highest votes. Small numerical uncertainty in Render vote predictions is evidently due to the lower number of input votes.

Figure 9 presents the relative importance of each feature, computed using Shapley values. This is the significance the model is giving to each individual feature, relative to the total significance. Note that the sum of significance for all features is 1. Figure 10 shows Shapley values for three selected points. This SHAP summary plot shows how each feature influences the model’s predictions, both in direction and magnitude. Points to the right (positive SHAP values) indicate that a feature value increases the prediction, while points to the left indicate that it decreases it. The color of each dot represents the actual value of that feature for a given sample—red/pink means a high feature value and blue means a low value.

4.2.2. Augmented Reality Vote Prediction Model

The Augmented Reality vote prediction model showed limited predictive performance, with predicted values clustering around the mean of the dataset. This is primarily due to the small size of the dataset, and the inherent subjectivity of Augmented Reality votes, which are strongly influenced by human visual perception and the specific tools or platforms used. As a result, the model predictions exhibit high variability and provide only a rough approximation of actual votes.

4.2.3. NFT Vote Prediction Model

For the analysis of NFT vote predictions, the Extra Trees Regressor was selected as the best-performing model. Its evaluation metrics indicate accurate predictions, with a low mean absolute error (0.022) and mean squared error (0.0011). The model also shows consistent predictions with a root mean squared error (0.0263).

The obtained results suggest that Area and Total Points play a dominant role in predicting NFT votes. According to the feature importance plot (Figure 11), Area contributes the most to the model’s accuracy, followed by Total Points. The SHAP summary plot (Figure 12). further confirms this relationship, indicating that higher values of these two features positively influence the predicted number of votes. In contrast, Elevation dispersion and Complexity show relatively low importance and a limited effect on the model output, implying that their contribution to vote prediction is marginal.

5. Discussion

5.1. Theoretical Contribution: Representation as an Active Parameter in Participatory Evaluation

Although architectural and artistic design competitions are widely examined in academic literature [48,49,50,51], the role of representational formats in shaping outcomes has received far less attention [77,78,79], especially when combined with participation of the broader public [15,16,17]. In most discussions, visualization is treated as a neutral medium for conveying design intent, rather than a factor capable of influencing aesthetic judgment and decision-making.

This research contributes to closing that gap by demonstrating that the choice of representational platform—static rendering, NFT-based animated representation, or Augmented Reality spatial visualization—directly shapes both who participates and how decisions are made. Different visualization formats attract different audience groups, each with distinct levels of digital literacy, familiarity with virtual environments, and perceptual habits. Therefore, representation functions not simply as a means of communication but as an active mediator that conditions the evaluative process itself.

This insight adds a new dimension to theoretical discourse on public architectural competitions: representation is not merely descriptive but performative, structuring participation and influencing the legitimacy and inclusivity of the competition outcome. Thus, the manner in which architectural proposals are shown cannot be separated from the ethics, epistemologies, and social dynamics of collective decision-making.

5.2. Comparative Participation Outcomes Across Platforms

The central aim of this research was to determine whether different digital platforms for participation lead to distinct participatory outcomes; in other words, to examine whether the very choice of platform influences the nature of public engagement and decision-making results.

In terms of engagement levels, participation rates varied substantially: 68.4% of the total participants engaged through the NFT platform, 20.7% through renderings, and only 10.8% via Augmented Reality representations. This distribution was somewhat unexpected, as the initial assumption anticipated that renderings—being the most familiar and visually accessible format—would attract the largest number of participants. Despite the fact that the survey was disseminated in an equivalent manner across all three modalities (through direct communication, email, and social networks), the NFT platform elicited the strongest response.

The findings presented in Section 4 suggest that this assumption may indeed hold true. Notably, one sculpture emerged as “winner” within two separate representational contexts—URBS#1 (corresponds to ID_0) in the NFT and Augmented Reality environment (Figure 13a,b). On the other hand, URBS#7 (corresponding to ID_20) stood out primarily in the static 2D render survey, attracting the most attention in that type of visual representation (Figure 13c). This indicates that differences between traditional visual representations and emerging technological modalities significantly influence participant preferences. Although further investigation is necessary to confirm and expand upon these results, the data obtained thus far demonstrate clear differentiation in audience behavior across platforms.

It is important to note that, in addition to differences in sculpture preferences across visualization types, variations were also observed with respect to the participants’ level of technical knowledge. Since the surveys related to the static render and Augmented Reality formats included a self-assessment question regarding technical proficiency, the results (Figure 14) for the static render ranged from basic (21.29% of participants) to professional (25% of participants), with the overall distribution among none, basic, intermediate, advanced, and professional categories being relatively balanced.

In contrast, the Augmented Reality survey results showed that none of the participants reported having no technical knowledge. The majority identified as advanced (36.4% of participants), while professional and intermediate levels were evenly represented (24.2% each). This finding suggests that the evaluation of sculptures in Augmented Reality environments tends to attract or be more accessible to technically proficient users. Such a trend is expected, considering that Augmented Reality interactions require a greater degree of familiarity with digital tools, mobile applications, and spatial navigation compared to traditional 2D or static 3D representations. Consequently, while Augmented Reality technologies offer enhanced immersion and engagement potential, their effective use still relies on a certain level of technical competence. This limitation raises concerns about the inclusivity of Augmented Reality -based participatory approaches, especially when addressing broader, non-specialized audiences. This aligns with previous studies dealing with the practical challenges of using mixed reality in different participatory contexts, suggesting that technical barriers can limit the inclusivity of Augmented Reality -based participation [80]. These findings collectively indicate that, despite their potential, Augmented Reality and mixed reality platforms still face usability and accessibility constraints that need to be addressed to ensure more inclusive public engagement and learning experiences.

This finding appears consistent with the central hypothesis, which proposed that each digital platform attracts different societal segments characterized by distinct degrees of digital literacy and technological familiarity. These variations, in turn, influence the evaluative criteria participants employ when making aesthetic or design-related decisions. The NFT platform, hosted on Objkt, appears to have drawn a community already accustomed to digital art environments—comprising artists, collectors, and digitally engaged audiences. Its guiding principle, “Art for everyone,” aligns with the participatory inclusivity that may have contributed to the higher number of responses. Furthermore, the dynamic and time-based nature of NFT presentations—offered as short videos rather than static images—likely enhanced their perceptual appeal and engagement potential compared to still renderings.

This interpretation is reinforced by the outcomes of the Regression Model Development and Feature Analysis, which revealed that Area and Total Points were the dominant predictors of votes within the NFT context. For render-based evaluations, Total Points emerged as the primary determinant. Conversely, the Augmented Reality model displayed limited predictive accuracy, with the predicted values clustering around the mean, indicating that no single parameter exerted a predominant influence on participants’ decisions. This suggests that perceptual variability within Augmented Reality experiences—dependent on numerous external factors—complicates consistent evaluative patterns.

5.3. Effects of Visualization Conditions and Environmental Factors

An additional finding of particular relevance is the significantly smaller number of participants willing or able to take part in the Augmented Reality-based survey. This limitation can be attributed to several interrelated factors, which may be grouped into three primary categories:

• Technological factors: Augmented Reality functionality was restricted to users with iOS devices and sufficient available storage space, thus significantly narrowing the potential participant pool. This technical dependency reveals ongoing accessibility constraints, as effective interaction with Augmented Reality environments often presupposes prior familiarity with digital tools and adequate technological infrastructure. Similar challenges have been identified in previous research on the implementation of mixed reality systems in participatory and educational settings, indicating that such technical prerequisites can restrict broader and more inclusive involvement [80].
• Environmental factors: The viewing environment—whether interior or exterior, natural or built, under varying lighting conditions—exerted a considerable influence on the perception of the Augmented Reality sculpture. Although participants were advised to view the objects outdoors, many likely did not do so, given the additional effort required, which further limited the quality and consistency of feedback.
• Aesthetic factors: While Augmented Reality presentations offer the most realistic approximation of real-world spatial perception, their quality depends heavily on user behavior and device conditions. Unlike renderings and NFTs, which are carefully curated and compositionally controlled, Augmented Reality experiences are contingent upon individual user settings and physical surroundings, leading to a high degree of variability in evaluative responses.

This variability likely contributed to the lower predictive accuracy observed in the Regression Model for Augmented Reality results.

5.4. Implications for Design Competitions and Future Research

Possible practical implications of this study include:

• The findings of this study underscore the necessity of reconsidering the conventional assumption that the visual format of submission materials is merely representational and does not directly impact the evaluation process. Instead, the results demonstrate that representational modalities function as active frameworks that shape audience composition, perception, and decision-making criteria. Therefore, the choice of visualization format in public architectural or art competitions cannot be treated as a neutral technical decision, but rather as a design parameter that influences the legitimacy, inclusiveness, and interpretive depth of participatory outcomes.
• For competition organizers, this presents a practical implication: the selection of representational media should be aligned with the intended public. If the aim is broad and non-specialist participation, platforms requiring high levels of digital literacy—such as Augmented Reality or NFT—may inadvertently restrict accessibility. Conversely, if the objective is to invite engagement from technologically experienced or digitally native audiences, emerging visualization technologies may expand both the depth and nuance of the evaluative process. Thus, decisions regarding representation should be considered as strategic rather than auxiliary.

Finally, several questions arising from this study indicate directions for future research. For instance, the observed tendency of some participants to consistently select the first sculpture presented raises the issue of potential primacy bias—a behavioral pattern that warrants additional controlled studies. Moreover, the perceptual variability introduced by Augmented Reality environments suggests that future methodologies might benefit from standardized viewing conditions or the development of guided Augmented Reality experiences. Further investigation into how participants interpret generative outcomes across media could also contribute to broader theoretical discussions on authorship, agency, and aesthetic cognition in digitally mediated public decision-making.

6. Conclusions

This study examined how emerging digital platforms, specifically Augmented Reality and Web3, can serve as instruments for broadening public participation and enhancing perceptual access to urban design proposals. An original algorithm was developed to generate nine abstract sculptural forms conceived for potential integration within public urban contexts. For each generated sculpture, a set of descriptive attributes was produced, forming the basis for both qualitative and quantitative analyses across different modes of visual representation.

The sculptural forms were presented to three distinct participant groups, each exposed to one type of visual representation: traditional renderings, Augmented Reality environments, and NFT-based digital representations within the Web3 platform. Collectively, these modalities articulate a continuum of representational conditions, from immersive spatial engagement to detached digital contemplation. They reveal how technological affordances mediate the phenomenological encounter between viewer and object, shaping not only aesthetic perception but also the participatory agency of the public. In this sense, the study positions digital representation not merely as a communicative tool but as an active framework through which social and perceptual relations are constituted.

Participants evaluated the artworks through a voting process, allowing each sculpture to accumulate a total number of positive votes within its visualization context. This procedure identified which forms were perceived as most successful in each environment. Comparative analysis of participant responses demonstrated that different modes of digital interaction significantly influence public perception, engagement, and the inclusivity of feedback processes.

The results indicate that the mode of digital mediation significantly shapes user perception and participation, as evidenced by differing audience preferences and engagement levels across NFT, rendering, and Augmented Reality environments. The NFT platform generated the highest participation and predictive consistency, reflecting the influence of digitally literate and art-oriented communities, while the Augmented Reality setting revealed greater perceptual variability and lower accessibility. This interpretation is further supported by the outcomes of the Regression Model Development and Feature Analysis, which identified Area and Total Points as dominant predictors of votes within the NFT context, and Total Points as the primary determinant for render-based evaluations. In contrast, the Augmented Reality model exhibited limited predictive accuracy, with values clustering around the mean—indicating that no single parameter had a decisive impact on participants’ choices.

Overall, these findings confirm that distinct digital platforms, both traditional and emerging, attract different user groups and foster specific evaluative behaviors, thereby mediating both perception and participation within socially sustainable design frameworks. This study argues that the choice and critical evaluation of the digital platform must precede the design of participatory processes; understanding how specific technologies mediate interaction and feedback is essential for ensuring that digital participation contributes meaningfully to social sustainability. Nevertheless, there remain several questions that this study only begins to open. For instance, the observed tendency of participants to select primarily the first sculpture presented raises further issues regarding perceptual bias and decision-making behavior. Such patterns require additional, more targeted research to determine whether they reflect technological influence, cognitive primacy effects, or other factors within participatory design contexts.

In emphasizing the need for reflective integration of emerging media, the research seeks to contribute to the ongoing discourse on how technology can advance more inclusive, transparent, and equitable approaches to urban and architectural design.