2.1 Character design and psychological responses

Character design in commercial contexts represents a complex interdisciplinary field integrating aesthetic principles, technological applications, and cultural frameworks to develop engaging entities across various media platforms (Miketić et al., 2018). In psychology, the term “character” has traditionally referred to internal personality traits and one’s relationship with the external world (Baudry, 1984), whereas in storytelling or data visualization, character refers to the identifiable agents that guide narrative understanding (Dasu et al., 2023). In design research, related concepts such as design frames are often ambiguously defined and context-dependent, requiring greater conceptual clarity to develop coherent cognitive models (Kelly and Gero, 2022). These varying definitions underscore the need to precisely define the notion of “character” within a given research context. In this study, we define a character as a visually rendered, anthropomorphic entity designed with the intention of eliciting user engagement and potential market appeal (Li et al., 2025). This definition excludes utilitarian avatars used for user representation (e.g., in games or metaverse environments), non-anthropomorphic mascots, and characters embedded exclusively within linear media narratives (e.g., film protagonists). Instead, we focus on characters that are visually self-contained and possess the potential to function as standalone commercial products, such as figurines, character-branded merchandise, and collectible digital assets. In doing so, our scope aligns with the character merchandising industry, in which visual design alone, rather than embedded narrative or interactive role-forms the primary basis of consumer engagement and valuation.

Within the domain of three-dimensional animation, character design occupies a central position in the production pipeline, with particular emphasis on developing distinctive stylistic approaches that embody specific cultural signifiers (Wu and Chang, 2021). In particular, marketing and advertising literature has extensively explored the effects of spokes-characters. These studies suggest that characters can significantly influence consumer responses, including perceived trustworthiness (Garretson and Niedrich, 2004), likability (Phillips and Lee, 2005; Chang, 2014), and purchase intent by fostering positive brand attitudes (Chang, 2014). Consequently, spokes-characters play a crucial role in shaping brand identity and fostering consumer-brand relationships (Callcott and Lee, 1995).

Psychological responses to character designs encompass both cognitive and affective components, which, while distinct, are interrelated and often occur simultaneously (Bitner, 1992). Cognitive responses involve rational information processing, such as evaluating a character’s sincerity, trustworthiness, and competence (Folse et al., 2012; Danovitch and Mills, 2014). These evaluations are particularly influential in building trust between consumers and brands (Garretson and Burton, 2005). In contrast, affective responses include emotional reactions such as sympathy, empathy, and emotional alignment with a character’s traits or experiences (Hoorn and Konijn, 2003; Schramm and Knoll, 2015; Park et al., 2019). These emotional connections enhance brand loyalty (Mac Innis and Folkes, 2017). Therefore, thoughtful character design and authentic representation are crucial in shaping consumer perceptions, fostering emotional engagement, and aligning with brand values.

Consumer perceptions of a character’s authenticity, originality, and relatability strongly influence their willingness to purchase branded products (Aaker, 1997; Kim, 2011). However, most prior studies have concentrated on characters as tools for enhancing corporate branding and advertising effectiveness rather than the characters themselves as marketable products. This emphasis overlooks a critical distinction: characters can function not only as branding instruments but also as standalone consumer goods, particularly in industries such as animation, games, and merchandise. Few studies have investigated how intrinsic design attributes, such as visual complexity, affect the commercial viability of characters as products in their own right. Our study addresses this gap by shifting the analytical focus from characters as peripheral branding elements to their central role as product offerings. By doing so, we aim to investigate how psychological responses toward a character design influence user-side evaluations, such as preference and expected marketability.

It is important to clarify that ‘expected marketability’ in this study refers to consumers’ perceived potential of a character design to succeed in the marketplace, based on subjective evaluations collected through surveys, rather than actual commercialization decisions made by producers or firms. While actual market performance involves a complex set of strategic and operational factors, our focus remains on how design features and psychological responses contribute to the perceived market potential of characters from the consumer's perspective. In particular, we propose that visual complexity plays a pivotal role in shaping both preference and expected marketability, thereby contributing to the literature on design aesthetics and consumer behavior, and offering practical implications for character commercialization strategies.

2.2 Visual complexity in character design

Visual complexity has been conceptualized through various frameworks. Prior studies have measured it based on factors such as the number of visible objects, degree of irregularity, level of detail, object dissimilarity, and asymmetry in arrangement (Orth and Crouch, 2014). The theory of visual complexity (Attneave, 1954; Donderi, 2006) suggests that images contain redundancy—non-randomness of visual information, including similarity, proximity, and continuity of features—where lower redundancy corresponds to higher complexity. In general visual stimuli, increased complexity can heighten arousal, attract attention, and improve memorability through peripheral processing (Petty and Cacioppo, 1986; Deng et al., 2009), but may also impose cognitive load and reduce clarity (Pieters et al., 2010).

Research in aesthetic preferences more broadly has similarly demonstrated that stimulus complexity follows an inverted-U relationship—where preference increases with complexity up to an optimal point, after which it declines (Delplanque et al., 2019). Empirical studies in visual art and music further validate this pattern (Chmiel and Schubert, 2017; Gordon and Gridley, 2013; Hekkert and Van Wieringen, 1990; Madison and Schiölde, 2017; Marin et al., 2016; Messinger, 1998). Advertisements typically combine multiple visual elements—such as products, brand logos, backgrounds, models, and text—to establish brand positioning and enhance appeal (Alden et al., 1999; Liu et al., 2020). Prior research has shown that the quantity, regularity, symmetry, and arrangement of these elements significantly affect consumer perception and behavior (Cox and Cox, 2002; Donderi, 2006; Peracchio and Meyers-Levy, 1994, 2005; Favier et al., 2019; Tong et al., 2022). This principle has made visual complexity a key characteristic in advertising research, particularly regarding its influence on consumer attitudes (Pieters et al., 2010; Wu et al., 2016; Kusumasondjaja and Tjiptono, 2019; Baek et al., 2023; Pleyers, 2024). This principle suggests that character designs must balance simplicity and complexity to maximize appeal while avoiding excessive detail that overwhelms consumers (Hekkert and Van Wieringen, 1990). However,a direct application of these findings to character designs is not straightforward.

In the context of character design, a single, anthropomorphic visual object imbued with narrative and emotional cues, these mechanisms interact with additional processes such as emotional resonance, narrative identification, and anthropomorphism (Epley et al., 2007; Horton and Wohl, 1956). For example, when a character’s form becomes too intricate, viewers may struggle to interpret facial expressions or attribute personality traits quickly, whereas moderate complexity can enhance perceived richness and foster parasocial engagement. In contrast to advertisements—where overall design composition matters—a character’s color scheme, edge definition, and texture complexity are more central in shaping consumer perception (Pieters et al., 2010; Wedel and Pieters, 2014; Xiao and Ding, 2014). The concept of feature complexity refers to the variation and richness of pixel-level visual information that the human visual system processes at an early stage (Palmer, 1999). This includes unstructured and low-level features of the image, independent of any semantic or narrative content. In character images, higher levels of color diversity, edge density (or the standard deviation of edge strength), and texture variation contribute to a greater sense of feature complexity.

Each of these complexity dimensions is expected to evoke distinct psychological responses. Specifically, color complexity may stimulate visual curiosity and promote cognitive engagement by increasing visual search and processing effort (Huhmann, 2003; Ales et al., 2020; Sun and Firestone, 2021; Kanuri et al., 2024), whereas edge complexity can influence the perceived sharpness and clarity of a character’s form, impacting the ease of recognition and emotional expressiveness (Queiroz et al., 2014; Kendall et al., 2016; del Valle-Canencia et al., 2022; Sunstrum et al., 2024). Texture complexity may enrich sensory detail and drive emotional engagement by conveying material qualities or symbolic associations (Whitaker et al., 2008; Djonov and Van Leeuwen, 2011; Jacobs et al., 2016). These cognitive (e.g., attention, processing fluency) and emotional (e.g., affective resonance, visual pleasure) responses serve as essential determinants of outcomes such as preference and perceived marketability. Accordingly, we conceptualize a pathway in which each type of complexity activates specific psychological reactions, which then influence consumer behavior. However, when feature complexity is too high, it can hinder a viewer’s ability to interpret or emotionally connect with the character quickly. Excessive visual detail may distract from the core identity or emotional expression of the character, reducing clarity and engagement (Pieters et al., 2010). Important visual cues, such as facial features, silhouettes, or symbolic elements, may become obscured by unnecessary visual noise. On the other hand, a moderate level of feature complexity can act as a positive peripheral cue, enhancing visual interest, stimulating arousal, and improving memorability (Deng et al., 2009).

The Elaboration Likelihood Model (Petty and Cacioppo, 1986; Lee and Yoon, 2025; hereafter, ELM) provides a relevant framework, as visual complexity, through dimensions such as color, edge, and texture, can act as peripheral cues. These cues attract attention and shape consumer attitudes without requiring detailed cognitive elaboration. Rather than conveying explicit arguments, character visuals often function as aesthetic or affective signals that influence user preference through visual appeal, perceived quality, or emotional resonance. Therefore, applying the ELM framework is justified when character exposure occurs under low motivation, low involvement, or passive viewing conditions (e.g., social media, mobile search, physical shelf), which are common across both digital and physical touchpoints (Shin et al., 2020). This is particularly evident in low-motivation environments, such as social media or mobile platforms, where users scroll quickly and have limited capacity for in-depth processing. In such contexts, character design functions within a low elaboration setting, in which consumers are unlikely to engage in central-route processing of product information. Even in offline settings, such as when consumers encounter characters through merchandise, packaging, or figurine displays, character design is often processed spontaneously and visually rather than analytically. However, character images with too little complexity may fail to capture interest or stand out from competing content, while those with excessive complexity may overwhelm viewers and impair recognition. This supports a non-linear, inverted-U-shaped relationship between feature complexity and viewer preference (Morrison and Dainoff, 1972). Also, drawing on the processing fluency framework, the visual complexity of character images is hypothesized to exhibit an inverted-U relationship with consumer preference (Berlyne, 1973; Reber et al., 2004): low to moderate complexity (i.e., low to moderate cognitive load) enhances processing fluency and aesthetic pleasure, while excessive complexity (i.e., high cognitive load) reduces fluency, consequently diminishing perceived beauty and preference.

Existing research has not adequately addressed how visual complexity functions within character design as a standalone feature, nor has it empirically examined its influence on perceived marketability. Furthermore, most studies rely on subjective evaluations rather than objective image-based indicators.

This study addresses these gaps by applying quantifiable measures of visual complexity to character images and analyzing their relationship with consumer preference and marketability. Building on these insights, this study defines visual complexity in character design across three key dimensions, color, edges, and texture (Kaur and Sohi, 2016; Pavithra and Sharmila, 2018; Overgoor et al., 2022). These dimensions capture the essential visual attributes that shape consumer perception and preference. Based on this foundation, we propose the following hypotheses:

Hypothesis 1a (H1a): The color complexity of the character design images has an inverted-U relationship with consumer preference.

Hypothesis 1b (H1b): The color complexity of the character design images has an inverted-U relationship with the expected marketability of merchandised goods.

Hypothesis 2a (H2a): The edge complexity of the character design images has an inverted-U relationship with consumer preference.

Hypothesis 2b (H2b): The edge complexity of the character design images has an inverted-U relationship with the expected marketability of merchandised goods.

Hypothesis 3a (H3a): The texture complexity of the character design images has an inverted-U relationship with the consumer preference.

Hypothesis 3b (H3b): The texture complexity of the character design images has an inverted-U relationship with the expected marketability of merchandised goods.

The conceptual framework of this study is presented in Figure 1

3.1 Data collection and variables

We collaborated with an online platform company that facilitates transactions between creators and firms of character design in South Korea and administered an online survey from two distinct samples: (1) 100 female respondents in their 20s and 30s, randomly recruited through a well-known domestic market research platform¹⁾, and (2) 144 students and graduates from arts-related departments at a major South Korean university (predominantly composed of women). This sampling approach reflects the shifting customer base of the character design market toward millennial and Gen Z women (cf. The study’s potential limitations related to sampling bias and the generalizability of the findings are discussed in Section 5.3).²⁾ Demographic information, including age, gender, and major, was collected from these participants. To assess potential heterogeneity between the two sample groups, we performed t-tests and chi-square analyses. The only statistically significant difference pertained to participants’ art major status, which was accounted for as a control variable in the panel regression model³⁾. For the survey, we selected 150 representative character designs from the Korean market. We adopted a stratified sampling approach to select a total of 150 character designs, divided into three tiers: (1) Tier 1: 50 randomly sampled character designs that are widely recognized by the public and considered to have relatively high commercial value (e.g., Super Mario, Pororo, Sailor Moon); (2) Tier 2: 50 randomly sampled character designs with some degree of public recognition and listed on the marketplace of the online platform company (e.g., Overaction Rabbit, Dora, Nutty); and (3) Tier 3: 50 randomly sampled character designs with minimal recognition and not listed on the marketplace of the online platform company (e.g., Ato, Ogu, Cupid Bear). Public awareness of each character design was operationalized by ranking their Google Trends search volumes in descending order at the time of data collection, after which the 150 characters were reclassified into three distinct awareness tiers. Each participant viewed 30 randomly selected character designs from the pool of 150 items and responded to 19 five-point Likert-scale questions regarding their psychological reactions to the images, focusing on preference and expected marketability (see Appendix A, and Section 5.3 for discussion of the limitation arising from measuring the psychological response variable with a single item). Based on these responses, we constructed 6,530 cross-sectional data points. To ensure consistency, all character images had minimal backgrounds, front-facing views, and no character overlap. They were standardized to 800 × 800 pixels, vertically aligned, and centered with background elements removed. These adjustments preserved original creative intent while maintaining comparability.

As discussed, character designs often serve as key elements in branding. Therefore, in Table 1

Moreover, we extracted pixel- and object-level variables to characterize the features of the character designs. We focused on three dimensions to measure visual complexity: color, edge, and texture (See Appendix B). Color complexity was represented by color distribution entropy. For edge complexity, we used the standard deviation of edge strength and the proportion of pixels occupied by edges derived from Canny edge detection features. Texture complexity was assessed using the Local Binary Pattern (LBP) method. We reduced the LBP vector to a single dimension, which accounts for over 95% of the variance. Higher LBP values indicate more complex textures. Table 2

Table 3

3.2 Model

To assess the impact of visual complexity on preference and expected marketability of character design images, we used the main independent variables: color, edge, and texture. As indicated in the literature review, we incorporated squared terms for the four complexity variables to account for non-linear relationships. Additionally, we included low-level feature variables—such as brightness, contrast, and exposure (luminance and color distribution)—as covariates. At the object level, we controlled local structure and color relationships using OpponentSIFT, a feature extraction method robust to illumination changes (Van De Sande et al., 2009; see Appendix C). To account for high-level image features, we incorporated the variables extracted through a deep learning-based Swin Transformer model (Liu et al., 2021; Gong et al., 2022; see Appendix D). As discussed, we included the psychological variables measured by the survey. Additionally, we accounted for demographic factors such as age, gender, region, and major. All continuous independent variables were mean-centered prior to analysis to facilitate interpretation and reduce potential multicollinearity, especially when including quadratic terms. Our primary models are expressed in Equations (1) and (2):

where $EdgeDensity_j$ represents the proportion of pixels occupied by edges in character design j, extracted via Canny edge detection; $EdgeStd_j$ denotes the SD of edge strength in character design j; $Entropy Hist_j$ refers to the Shannon entropy of the red, green, and blue intensity values of character design j, represented as a one-dimensional directionless variable; $LBP_j</italic-math>$ captures the texture feature variable derived from the LBP vector, reduced to a single dimension using principal component analysis (PCA); and $LLF_j$ represents the low-level feature vector of character design j, including brightness, exposure, and contrast. $Edge Avg_j$ indicates the average edge strength in character design j, extracted using Canny edge detection; $OpponentSIFT_j$ accounts for color information; and $Swin Transformer_j$ captures local and global high-level feature variables of character design j, derived using a Swin Transformer model. Furthermore, $Psycho Vars_{ij}$ denotes the psychological response variables of user i toward character design j. $Prior Knowledge_{ij}$ represents user i&rsquo;s prior knowledge of character design j (i.e., creator recognition and character recognition variables). $X_i$ is a matrix representing the demographic characteristics of user i (i.e., a matrix consisting of the vectors $gender_i$, $age_i$, $region_i$ and $major_i$). $\mu_i$ represents a fixed effect for user heterogeneity, while $\pi_j$ represents a fixed effect for character design heterogeneity. Dependent variables include $Preference_{ij}$ (user&rsquo;s preference for character design j), and $Exp Marketability_{ij}$ represents user i&rsquo;s expected marketability of character design j. All independent variables involving squared terms were mean-centered to ensure the interpretability of linear coefficients in the quadratic regression models (Aiken et al., 1991).

To further examine the impact of unique characteristics, including visual complexity, influence on the likelihood of actual commercialization, we constructed a logistic regression model at the character design level, as shown in Equation (3).

where $Commercial_j$ represents the actual commercialization status of character design j.⁴⁾ $\pi_j$ represents a fixed effect for character design heterogeneity. Characters perceived by consumers as having high commercial potential are more likely to be selected for commercialization through corporate market research. Therefore, the mechanism of consumer response to perceived complexity, as discussed above, can similarly apply when the actual commercialization decision is used as the outcome variable.

4.1 Results

Table 5

⁴ The commercialization status of the character designs was determined through keyword search results on major Korean e-commerce platforms. These include Coupang, Naver Shopping, G-market, Auction, and 11st. This information was validated further using commercialization history data from the character design trading platform. We were able to identify a total of 52 commercialized characters.

Additionally, the negative regression coefficient of $(LBP)^2$ In the preference model, it indicates that the texture complexity of a character design image, represented by LBP, also exhibits an inverted-U relationship with consumer preference (i.e., H3a is supported). This implies that consumer preference peaks at an optimal level of texture complexity. However, no consistent results were observed regarding the relationship between texture complexity and expected marketability.

Regarding psychological responses, attributes related to symbolism, prominence, and target specificity did not significantly affect preference but positively and significantly impacted expected marketability. This indicates that consumers perceive character design images with a clear target audience, high memorability, distinct visuals, and symbolic attributes as having higher expected marketability. Additionally, prior knowledge of character design, including character or creator recognition, did not significantly influence preference or expected marketability.

Table 6

Table 7

4.2 Supplementary analysis

We applied K-means clustering to assess the relationship between visual features, psychological response variables, and commercialization outcomes. Clustering variables included color, edge, texture complexity measures from the main analysis, edge intensity-related variables, contrast, brightness, and exposure (i.e., “LuminanceSkew”), along with other low-level features and psychological response variables. Based on the elbow method, the optimal number of clusters was determined to be three. The relationship between these factors and the binary commercialization outcome $(Commercial_j)$ was examined.

Table 8

5.1 Theoretical implications

This study is among the first to examine the impact of character design image complexity on marketability. We identified distinct effects by categorizing visual complexity into color, edge, and texture dimensions. Additionally, we highlighted differences in perceived commercialization factors between individual consumers and corporate users. The results suggest that visual complexity is a critical consideration at the corporate (e.g., licensees) level, and factors such as contrast and maximum edge strength, which indicate clear object representation, directly influence their commercialization decisions.

In marketing, prior research on visual complexity and consumer responses has primarily focused on advertising and product packaging. This study expands the scope by applying these concepts to character designs as commercialized products. By integrating machine learning-based image analysis and low-level feature extraction, we control for intrinsic image characteristics while accounting for individual cognitive heterogeneity. These methodological advancements contribute to the growing literature on visual complexity, consumer perception, and character design marketability.

5.2 Managerial implications

This study underscores the importance of visual elements in corporate decision-making. As character designs are developed primarily for commercialization—in-house or through licensing—understanding their influence on consumer preference and marketability is essential for revenue generation. Although companies invest substantial resources in enhancing character design designs, excessive complexity may lead to adverse consumer reactions. This study highlights the need for a data-driven, market-aware approach to character design and licensing strategies.

These findings suggest that design teams should adopt a balanced approach to visual complexity, leveraging consumer insights early in the design process to avoid over- or under-designed characters. Marketing and product managers can benefit from using quantifiable visual indicators, such as edge intensity (e.g., Edge Avg), color variation (e.g., Entropy Hist), and texture clarity (e.g., LBP), to evaluate and refine character designs before market launch. Furthermore, firms utilizing generative AI or design automation tools should integrate complexity thresholds into their algorithms to align visual output with consumer preferences. In the context of licensing, companies can enhance pricing strategies by assessing how design features influence perceived value. Overall, a systematic understanding of visual complexity enables companies to better manage design portfolios, reduce market risk, and improve the effectiveness of character-based branding and merchandising efforts.

Moreover, the cluster analysis leveraging character complexity features extracted via machine learning offers concise, practice-oriented insights for firms and practitioners pursuing character commercialization, specifically, which market segments to target and how to target them. For example, when organizations develop characters to meet immediate the companies' marketing needs, customers' psychological responses are often insufficiently considered; this gap can be remedied by explicitly incorporating such considerations into character design. Likewise, when creators or firms design characters independently, a strategy of conducting market research to identify features likely to elicit favorable psychological responses, and integrating those features into the design process, may be warranted.

5.3 Limitations and future research

This study has several limitations that call for additional research. First, the character design samples were limited to the South Korean market. As cultural factors influence character design evaluations, future research should explore global character designs to improve generalizability.

Second, the participant sample consisted mainly of South Korean women in their 20s and 30s, including a large proportion of art students and graduates. While this reflects the growing influence of younger female consumers in character-driven markets, the findings may not generalize to other demographics such as men, older consumers, or those from different cultural backgrounds. In addition, unmeasured factors, such as aesthetic sensitivity or design-related motivation, may have influenced responses. Broader and more diverse samples are needed in future research to improve external validity.

Third, the study included character designs that had already been commercialized and available on the market as part of the survey. This may have led to overestimating expected marketability among survey participants, which introduced bias into the estimation results. Although we controlled for prior knowledge of the character design or its creator in our model, this approach may not entirely eliminate potential biases. To mitigate such issues, it is necessary to adopt a blind test design focusing on non-commercialized characters or to refine the level of control over recognition-based variables. Additionally, we measured the psychological variable using a single item, constituting a potential threat to measurement validity. Future research could more robustly address potential reliability and validity concerns by employing multi-item measures (e.g., Cronbach’s alpha test).

Fourth, the regression models identified several statistically significant predictors of character preference and perceived marketability. Although some coefficients, particularly for psychological reaction variables, were modest in size (e.g., 0.01 to 0.05), they nonetheless point to consistent patterns worth noting. These findings highlight the importance of considering both statistical significance and effect size when interpreting results, and future research could further examine how these dimensions interact in shaping consumer responses.

Finally, the study did not account for additional commercialization factors, such as character design popularity in mass media and word-of-mouth effects on social media. Given the unpredictable nature of entertainment product success, a more systematic examination of external market dynamics would strengthen the findings. Future research should analyze these factors to offer a more comprehensive framework for character design commercialization.

Appendix B. Visual Complexity Measures of Character Design

Appendix B. Visual Complexity Measures of Character Design