In the realm of cultural expression, the classification of creative works—such as films, music, and literature—plays an essential role in shaping both audience perception and scholarly analysis. Yet, current systems often rest on genre definitions that shift over time, leading to distortions in how cultural products are understood and valued. A groundbreaking new study led by Demetrius Lewis, an assistant professor of management at the University of California, Riverside, reveals the significant impact of what he terms “retrospective bias” within genre classification systems. By leveraging advanced artificial intelligence techniques, Lewis and his colleagues have developed an innovative computational tool that reconstructs the contemporary cultural context of historical works, thereby correcting genre misclassification and offering deeper insights into cultural innovation.
Genres serve as foundational taxonomies that organize the vast universe of creative output. They guide studios in marketing strategies, inform streaming platforms’ content recommendation algorithms, and underpin academic studies of performance, trends, and innovation. For consumers, genre labels provide a heuristic to navigate the enormous volume of available works. However, the dynamic nature of genre definitions means that retrospective labeling often fails to capture the original intentions or audience reception of a work, which were conditioned by the cultural context at the time of release.
Lewis emphasizes that the definition of what constitutes a specific genre evolves substantially across decades. A horror film from the 1980s may fundamentally differ in tone, theme, and audience expectations when compared to its 21st-century counterparts. This temporal fluidity in categorization means that using contemporary labels to classify older works injects systematic errors, obscuring nuanced relationships between how these works were positioned originally and how they were received. This temporal misalignment is what Lewis and his co-authors classify as retrospective bias.
The ramifications of this bias extend beyond the academic realm into the lived experience of creators and audiences alike. Creative works that defy simple categorization by spanning multiple genres often embody innovation and boundary-pushing artistry. Yet, they risk being undervalued because existing genre frameworks penalize complexity and hybridity. According to Lewis, products that straddle too many categories may become challenging for audiences to interpret, leading to diminished reception and lower ratings, a phenomenon that constrains creative experimentation.
However, Lewis notes that the penalty for genre-crossing works diminishes over time, as audiences acclimate to novel genre combinations. Historical examples reveal how genres expand by assimilating innovative elements initially perceived as unconventional or perplexing. For instance, the fusion of acoustic folk and electric rock by iconic 1960s musicians like Bob Dylan and The Byrds eventually crystallized into the folk-rock genre, demonstrating how genres co-evolve alongside artistic innovation.
Similarly, genre evolution can be witnessed in the cinematic realm with films such as the 1976 horror classic “Carrie.” This film broadened the horror genre by intertwining supernatural terror with intensely personal narratives addressing puberty, identity, and social alienation, reframing genre boundaries and expanding audience expectations. Such examples underscore how genres are not static definitions but living constructs reflecting ongoing cultural experimentation.
To confront the complexities of genre evolution and retrospective bias, Lewis and his collaborators designed a computational framework harnessing the power of large language models (LLMs). This AI-driven tool analyzes textual and contextual genre data to harmonize genre classifications across multiple temporal layers. Essentially, the system translates between past and present genre taxonomies, revealing how cultural products would have been understood at the time of their initial release rather than through the lens of contemporary classification.
The methodology involves prompting the GPT-based large language model to assign genre labels twice for each analyzed work: once based on its current (or contemporary) knowledge base, and once confined to the cultural context available during the work’s original release year. By comparing these dual classification outputs, the system exposes shifts in genre definitions and evaluates their effects on how creative works have been categorized and interpreted over time.
This dual-contextual approach mitigates retrospective bias, offering a more faithful representation of a creative work’s original cultural significance. For creators, it ensures that boundary-defying projects receive equitable recognition, reducing the risk that innovation is penalized by rigid, anachronistic genre boxes. Lewis highlights that many artistic breakthroughs we now take for granted were once revolutionary and often misunderstood innovations that expanded genre horizons.
Audience benefits are equally profound. More accurate genre classifications can refine content discovery mechanisms, enabling viewers and listeners to appreciate the full intent and scope of creative works without oversimplification. This enriched contextualization enhances cultural literacy, deepens engagement, and fosters a more sophisticated understanding of artistic expression across eras.
The study’s insights have vital implications for industries reliant on genre tagging—from film studios and streaming services to music platforms and academic research institutions. As culture continues to evolve and genres proliferate and blend, adapting classification systems to reflect temporal dynamics is imperative for preserving the integrity and richness of creative expression.
Lewis and his team’s pioneering research, published in the Academy of Management Discoveries, exemplifies how advanced computational modeling and AI can intersect with cultural theory to tackle longstanding interpretive challenges. Their work paves the way toward a more nuanced and equitable cultural landscape where innovation is accurately recognized and historical works receive their rightful contextual framing.
Ultimately, this research challenges us to reconsider the seemingly immutable nature of genre and classification. By revealing the fluidity and evolution of genre systems, Lewis invites creatives, consumers, and scholars to engage with culture through a lens that honors its complexity and change, enriching our collective experience of art and storytelling.
Subject of Research: Not applicable
Article Title: Accounting for Retrospective Bias in Classification Systems of Cultural Products
News Publication Date: 30-Jan-2026
Web References: http://dx.doi.org/10.5465/amd.2024.0261
References: Lewis, D., Negro, G., & Guler, I. (2026). Accounting for Retrospective Bias in Classification Systems of Cultural Products. Academy of Management Discoveries.
Image Credits: UC Riverside
Keywords: retrospective bias, genre classification, artificial intelligence, large language models, cultural products, genre evolution, creative innovation, computational modeling, cultural context
