04/22/2026 | Press release | Distributed by Public on 04/22/2026 16:00
Featuring gory attacks by bloodthirsty vampires, one may be quick to categorize "Sinners" as a horror movie. That classification, however, may not be fair to the artists who created it. In "Sinners," the creators cleverly use horror as a metaphor for violent racism in the Deep South during the early 20th century, making the film just as much a period drama as a horror film.
Demetrius LewisNew research led by a UC Riverside professor shows this mismatch is not just a matter of taste - it is a measurable problem that can distort how creative work is understood, evaluated, and even discovered.
Demetrius Lewis, an assistant professor of management in the School of Business, has developed an artificial intelligence-based approach that helps correct these distortions, offering benefits for both creators and audiences.
Genre shifting matters because much of what people know about films, music, and other cultural products depends on how they are categorized. Studios use genres to market films. Streaming platforms rely on them to recommend content. Researchers use them to study innovation and performance. And viewers rely on them to discover works.
But genre labels applied years after a work is released are based on contemporary definitions that may not reflect how the work was originally intended or received, Lewis said.
"Genres change over time. What it means to be a horror movie in 1980 versus today is very different," he said.
Lewis and his co-authors argue that this "retrospective bias" can lead to systematic misunderstandings. Their study, published in the journal "Academy of Management Discoveries," finds that using modern genre labels to analyze older works can obscure important relationships between how a work was positioned and how audiences responded to it.
For creators, the consequences can be significant. Works that span multiple genres - often a hallmark of creativity - can be penalized because they do not fit neatly into established categories.
"There's this idea that when a product spans too many categories, it becomes harder for audiences to interpret," Lewis said. "And that can lead to lower ratings."
This penalty, however, is not fixed, he said. As audiences become more familiar with new combinations of styles, what once seemed confusing can become standard - even celebrated. Over time, genres expand, absorbing the very innovations that once challenged them.
The creators of the movie "Carrie" and recording artist Bob Dylan are acclaimed for creative work that went beyond the boundaries of the genres of the time. (Getty Images)Consider how 1960s musicians like Bob Dylan, Simon and Garfunkel, and The Byrds blended acoustic folk and electric rock genres to create what would become the popular folk-rock genre.
Similarly, the 1976 movie "Carrie," expanded the horror genre by exploring how a painfully shy teenage girl navigates puberty and identity as she deals with social rejection and bullying from her school peers.
"Genres that grow more complex over time tend to become broader," Lewis said. "They take on additional characteristics, and that reflects experimentation or innovation on the part of producers."
To capture the evolution of genres, the research team developed a computational tool that uses large language AI models to harmonize genre classifications across time. The system analyzes how genre definitions shift and then translates between past and present classification systems.
In effect, the tool reconstructs how a film, album, or other cultural product would have been understood when it first appeared, rather than how it is labeled today, providing nuance that situates the consumers within the cultural context of the time.
To build the tool, Lewis and his co-authors prompted the GPT large language model, asking it to assign genre labels for creative works twice - once using the model's current knowledge and once using only the cultural context that would have been known in the year the work was released. Comparing the two sets reveals how the classification and the overall genre system have shifted over time.
For creators, the approach offers a fairer assessment of their work. By accounting for how genres evolve, the tool reduces the risk that boundary-pushing projects will be misclassified or undervalued.
"Some of the things that were innovative when they first happened, we take for granted now," Lewis said.
For audiences, more accurate genre classifications can improve how people discover and interpret creative works. Instead of being guided by labels that may oversimplify or misrepresent a film, viewers can gain a clearer sense of its full scope and intent.
The title of Lewis' paper is "Accounting for Retrospective Bias in Classification Systems of Cultural Products." Its co-authors are Giacomo Negro of Emory University and Isin Guler of the University of North Carolina at Chapel Hill.
Lewis believes that better tools - particularly those powered by AI - can help strike a balance.
"Our goal is to better capture what these creative products really are," he said, "and how people experienced them when they first came out."