Open research questions in Aesthetic Perception and Analysis
47 unresolved questions extracted from the limitations and future-work sections of 1,600 Aesthetic Perception and Analysis papers in our library. Each links back to the study that raised it.
What the literature leaves open
Limitations arose due to the controlled lab setting of our study, as participants were encouraged to take photos for exploration. This differs from contexts in which participants would naturally take photos, which may be more sporadic in timing and more signif- icant in the captured environment. When we take the view that photographs are for capturing the moment for future remembrance, then participants sometimes indicated a need for more time to pro- cess their photos and reflect on them. Future work could extend towards a full, longitudinal in-the-wild study in which the system is incorporated naturally into one’s photographic routine over a longer period. Studying the extended use of the UnReality Cam- era could also help minimize the novelty effects from first-time use. This could also support a more detailed examination of the actual photographic artifacts themselves, including how users might per- ceive and experiment with different types of spoken prompts and the different styles. The perception of different generative styles could be examined in more detail as a factor of authorship and control; this factor was underexplored in our findings. Furthermore, participants were from a similar demographic range, and few had advanced experience with photography. While we found this adequate for an initial speculative exploration, we highlight how the use of photography as an art form rather than simply capturing the moment might tie more towards professional photographers; this could be an interesting population to purpo- sively sample in the future, as we hypothesize that professionals might be able to offer a more detailed account of the latter moti- vation, as well as the potential for stronger baseline comparison against their existing practice (especially if they have extensive instant photography experience). While we did not stratify our data analysis based on photography experience, our informal obser- vation of the data showed that people with advanced photography experience were more articulate at expressing their expectations, but still showed a range of responses from being frustrated at mis- alignment to being open to serendipitous interpretation. Future work could extend to a broader demographic, incorporating quan- tifiable work with statistical measures and causal inferences based on the impact of generative AI across different experiences. This could involve examining the frequency and valence of responses to provide more targeted takeaways. Our work uses a specific prototype in a probing study to address broader questions regarding people’s relationships with AI, percep- tions of stochastic generation in creative domains, and the impact of experiential dimensions such as temporality and materiality. However, the UnReality Camera deliberately constrained certain factors of freedom and control, leading to an unexplored design space. Based on our foundation in understanding design, alternative designs could shift their focus to more deeply addressing usability of our system through, e.g. the customizability options desired by many participants that may impact alignment in co-creation (hav- ing the option to quickly iterate on an image before printing, being able to generate multiple images, incorporating memory of user preferences over time, etc.). Accessibility and the context of use are other explorations that would be valuable. For instance, the use of speech as prompt input may not work for all users (e.g. due to privacy or during situational impairment [86]), and future work can look into alternative inputs and their effects on co-creation.
Re-Envisioning Instant Photography using Generative AI: An Exploratory Design Probe Using the UnReality Camera · 2026 · DOILooking ahead, future research should examine how socio-cultural factors influence affective involvement and whether the short-term emotional changes seen in EEG translate into long-term psychological outcomes.
The paper proposes a collaborative interaction model based on prompt-based input and iterative refinement but provides no empirical analysis of failure modes, such as cases where AI-generated variants diverged significantly from artist intent or where the 500% increase in stylistic options created decision paralysis or diminishing returns in creative exploration.
ARTIFICIAL INTELLIGENCE AS A CREATIVE COLLABORATOR IN CONTEMPORARY VISUAL AND PERFORMING ARTS · 2026 · DOIThe comparative workflow analysis focuses on creation time and stylistic diversity metrics but does not address how AI-assisted pipelines perform on domain-specific artistic constraints such as maintaining conceptual coherence across multi-piece series, respecting artistic style consistency, or integrating client feedback across iterations in commercial visual arts contexts.
ARTIFICIAL INTELLIGENCE AS A CREATIVE COLLABORATOR IN CONTEMPORARY VISUAL AND PERFORMING ARTS · 2026 · DOIThe paper reports artist satisfaction scores (8.6/10 for AI-assisted, 7.1/10 for human-only) and visual quality ratings (8.3 vs 7.4) but does not investigate whether the perceived quality improvements stem from the AI model's technical capabilities (texture composition, color balance, spatial equilibrium) or from the expanded creative search space enabling better artist selection among options.
ARTIFICIAL INTELLIGENCE AS A CREATIVE COLLABORATOR IN CONTEMPORARY VISUAL AND PERFORMING ARTS · 2026 · DOIWhile the paper demonstrates that artists maintained creative control through iterative prompt refinement and feedback mechanisms, the specific cognitive and decision-making processes by which artists evaluate and select among the 18 AI-generated stylistic variants (versus 3 manually created) have not been analyzed through think-aloud protocols or eye-tracking studies.
ARTIFICIAL INTELLIGENCE AS A CREATIVE COLLABORATOR IN CONTEMPORARY VISUAL AND PERFORMING ARTS · 2026 · DOIThe evaluation metrics in Table 4 and Table 5 were collected from a curated collection of digital artworks with unspecified sample size and artist demographics; the generalizability of the 72% time reduction and 500% style variant increase across different artistic domains, skill levels, and AI model architectures requires validation with larger, more diverse participant cohorts.
ARTIFICIAL INTELLIGENCE AS A CREATIVE COLLABORATOR IN CONTEMPORARY VISUAL AND PERFORMING ARTS · 2026 · DOIThe study evaluated AI-assisted creative workflows only within digital visual arts using generative diffusion models; the extent to which these findings apply to performing arts (dance, theater, music composition) with different AI architectures (transformer-based models, reinforcement learning systems) remains unexamined.
ARTIFICIAL INTELLIGENCE AS A CREATIVE COLLABORATOR IN CONTEMPORARY VISUAL AND PERFORMING ARTS · 2026 · DOIThematic metadata trend analysis is mentioned as identifying shifts in subject matter description over time, but the paper does not specify the controlled vocabulary or ontology used for theme annotation, nor does it address how semantic ambiguity and anachronistic terminology in historical metadata affects computational trend detection accuracy.
The integration of user interaction data from augmented reality and virtual exhibitions (navigation patterns, interaction duration, engagement metrics) with cultural analytics is mentioned as future work, but lacks specification of which engagement indicators correlate with aesthetic preference, or how immersive setting data should be extracted and normalized across heterogeneous digital platforms.
Real-time cultural analytics on continuously generated visual content from social media and digital exhibitions is proposed as a future direction, but the paper lacks specification of temporal resolution requirements, streaming data preprocessing pipelines, or methods for distinguishing emerging micro-trends from noise in high-velocity cultural datasets.
The paper proposes multimodal cultural analytics combining visual images with textual metadata and natural language processing on curatorial narratives, but provides no concrete implementation details, evaluation framework, or case studies demonstrating how text-image fusion specifically improves aesthetic trend detection beyond unimodal visual analysis alone.
The color histogram analysis demonstrating increased color variance during Impressionism and Post-Impressionism is presented without specifying the computational metrics used (e.g., color entropy, saturation ranges, luminance distributions) or dataset composition. Reproducibility requires explicit definition of color statistical measurements and the specific art collection parameters used.
The paper applies cluster analysis and AI-based visual similarity classification to historical art movements (Renaissance, Baroque, Impressionism, Modernism) but does not specify validation metrics or comparison against art historian expert classifications. Validation of whether computational clustering actually corresponds to established art historical periodization requires systematic benchmarking against human expert annotations.
Table 2 shows that final artwork selection has 86% perceived authorship score (highest human influence), but the paper does not examine whether this perception varies across different demographic groups, artistic expertise levels, or cultural backgrounds. The generalizability of authorship perception findings across diverse human evaluators requires investigation.
The paper proposes solutions for dataset ethics including 'curated databases' and 'better documenting the source of training,' but does not specify the technical infrastructure, metadata standards, or database schema requirements needed to implement such ethical AI training practices at scale for art generation systems.
The paper states that human control is highest in prompt-based AI art systems (78.1% control level) but does not systematically investigate how different prompt engineering strategies, prompt specificity levels, or prompt length variations affect the final perceived authorship scores and creative output diversity across different generative model architectures.
The paper discusses the ethical issue of AI systems reproducing stylistic features of existing artworks without crediting original artists, but does not propose or test technical mechanisms for tracking and attributing source artworks in AI-generated outputs. There is a need to develop and evaluate traceback methods that can identify which training artworks influenced specific generated works, particularly in style adaptation processes (46.8% AI contribution).
The paper presents a comparative performance analysis of AI art platforms (Table 3) showing diffusion model generators achieve 92.8% visual quality and 88.6% creativity index, but does not explain the evaluation criteria, judges/raters used, or sample size for these metrics. The validation methodology for these creativity and quality indices across GAN-based, diffusion, and prompt-based systems requires explicit definition and reproducibility documentation.
The paper identifies cultural bias in generative AI training datasets but does not specify which artistic traditions, geographical regions, or cultural styles are underrepresented in current datasets. Future work should conduct a systematic audit of training datasets used in GANs, diffusion models, and prompt-based AI art systems to document which cultural representations are imbalanced and by what magnitude.
The paper quantifies human and AI contributions across creative components (Table 2) but does not establish empirical validation methods to measure these percentage allocations objectively. There is a need for standardized metrics and experimental protocols to measure the creative contribution percentages (concept development 78.5%, style adaptation 46.8%, dataset preparation 64.7%) across different AI-generated art systems and human-AI collaboration workflows.
The extent to which AI-generated images can be used to explore cultural stereotypes and mental representations of the nursing profession has not yet been studied.
In other words, because of several guest-edited special issues of VAR in a row, we did not have opportunity to reward the Elliot Eisner Doctoral Research Award winners with their promised publication in the open issue of VAR in the year of their award.
In fact, VAR is devoting this whole open issue to showcasing their greatness, or “the quality or state of being important, notable, or distinguished,” and also “the quality or state of being powerful or intense” (https://www.
The principal contribution of Barbara Carnevali’s study is to draw attention to an area of aesthetics that is at once all-too-familiar and, perhaps for that very reason, still underappreciated and underexplored; although the lack of appreciation and interest is also due, as Carnevali argues, to long-standing resistance.
Most-cited papers in Aesthetic Perception and Analysis
- Ten years of a model of aesthetic appreciation and aesthetic judgments : The aesthetic episode – Developments and challenges in empirical aesthetics · British Journal of Psychology · 2014 · 406 citations
- Generative artificial intelligence, human creativity, and art · PNAS Nexus · 2024 · 363 citations
- Does human–AI collaboration lead to more creative art? Aesthetic evaluation of human-made and AI-generated haiku poetry · Computers in Human Behavior · 2022 · 191 citations
- Amplifying the Mind's Eye: Sketching and Visual Cognition · Leonardo · 1990 · 187 citations
- Humans versus AI: whether and why we prefer human-created compared to AI-created artwork · Cognitive Research Principles and Implications · 2023 · 162 citations
- Icon Identification in Context: The Changing Role of Icon Characteristics With User Experience · Human Factors The Journal of the Human Factors and Ergonomics Society · 2007 · 120 citations
- How Deepfakes and Artificial Intelligence Could Reshape the Advertising Industry · Journal of Advertising Research · 2022 · 120 citations
- Aesthetic preference for art can be predicted from a mixture of low- and high-level visual features · Nature Human Behaviour · 2021 · 108 citations
- The Primacy of Beauty in Judging the Aesthetics of Objects · Psychological Reports · 2004 · 103 citations
- Assessment and Therapeutic Application of the Expressive Therapies Continuum: Implications for Brain Structures and Functions · Art Therapy · 2010 · 101 citations
Most recent work
- When algorithms plate: Can AI-generated aesthetics rival reality? · International Journal of Gastronomy and Food Science · 2026
- Neural dynamics of aesthetic appreciation: fNIRS evidence from poetry reading · NeuroImage · 2026
- "There is Beauty in the Small": A Study of Visual Artists’ Impressions of Small Data and Model Crafting Approaches to Creative AI Work · 2026
- Exploring the emotional experiences of implementing GenAI-assisted digital multimodal composing: an autobiographic self-study · Innovation in Language Learning and Teaching · 2026
- To Use or Not to Use Artificial Intelligence (AI) in Art Design? A Multigroup UTAUT2 Model Based on Motivation and Engagement Profiles · Journal of Educational Technology Development and Exchange · 2026
- Artificial Intelligence, The Hedonic View of Appreciation and the Threat of Appreciative Deskilling · The Journal of Aesthetics and Art Criticism · 2026
- ARTIFICIAL INTELLIGENCE AS A CREATIVE COLLABORATOR IN CONTEMPORARY VISUAL AND PERFORMING ARTS · ShodhKosh: Journal of Visual and Performing Arts · 2026
- ARTIFICIAL INTELLIGENCE-GENERATED ART AND THE QUESTION OF AUTHORSHIP · ShodhKosh: Journal of Visual and Performing Arts · 2026
- CULTURAL ANALYTICS IN VISUAL ARTS: USING DATA SCIENCE TO UNDERSTAND AESTHETIC TRENDS · ShodhKosh Journal of Visual and Performing Arts · 2026
- Designing with the Brain in Mind: A Neuroaesthetic Pedagogy Framework · International Journal of Graphic Design · 2026
Find a gap in your own Aesthetic Perception and Analysis sub-topic
This page shows what the Aesthetic Perception and Analysis literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.
Open the Research Gap Finder →