Skip to main content

Verified by Psychology Today

AI Spots Success Formula for Artists, Directors, Scientists

Northwestern University’s AI identifies pattern onset of career “hot streaks."

ElisaRiva/Pixabay
Source: ElisaRiva/Pixabay

A new study by Northwestern University researchers published in Nature Communications uses artificial intelligence (AI) deep learning to provide insights and identify common patterns on career hot streaks across artistic, cultural, and scientific realms.

Successful artists, critically-acclaimed film directors, and Nobel Prize-winning scientists seemingly share something in common despite being in different professions—they have all had periods of career marked by “hot streaks”—successful bursts where their work was significantly better than their typical performance.

But why does this happen and is there a formula to explain it? In pursuit of understanding why certain creative talent experience hot streaks in their careers and not others, scientists at Northwestern University apply the predictive capabilities of artificial intelligence machine learning.

“In this paper, we constructed large-scale datasets for creative products from a wide range of disparate sources, including images of artworks for artists, film plots and casts for directors, and publication and citation information for scientists, along with their impact measures of auction prices, IMDb ratings and paper citations, respectively,” the researchers wrote.

To identify the metrics for success, the study used data from more than 2,100 artists, 4,300 directors, and 20,000 scientists.

“We developed computational tools from deep learning and network science and learned high-dimensional representations for these creative products, allowing us to trace an individual’s career trajectory on the underlying creative space around the beginning of a hot streak,” the researchers wrote.

To create high-dimensional representation of artworks, the researchers used a convolutional neural network algorithm for image recognition, specifically a pre-trained VGGNet algorithm, in combination with a fully connected neural network to classify art style labels in the dataset.

“We apply our deep neural network to the career outputs of each artist in the dataset, and then use principal component analysis for dimensionality reduction to generate a 200-dimensional embedding of each artwork,” wrote the researchers.

Then the team developed high-dimensional representations of films using the film’s plot and casting information. The scientists trained word embeddings in the plot description to learn a 100-dimensional text representation of a film from the co-occurrence of words. They used DeepWalk, a type of graph neural network (GNN), to get a 100-dimensional casting vector for each film. By linking the vectors for plot and cast, the researchers created a 200-dimensional embedding space to represent films.

“Despite the myriad factors that may affect the artistic and financial success of a film, ranging from the screenplay to acting, we find that the learned high-dimensional representation can successfully predict film genre with an accuracy of 0.948,” the researchers reported.

For the analysis of scientists, the researchers used AI to find patterns in the various research topics from the papers cited within a researcher’s published paper using the publication records of scientists from Web of Science and Google Scholar.

“The trade-off between exploration and exploitation — and its relationship to creativity and learning — has been discussed extensively across a broad set of disciplines, ranging from computer science, to psychology, to neuroscience, to computational social science, to strategic management and organization theory,” wrote the researchers.

The scientists looked at the diversity of the work produced before and during the bursts of hot streaks and compared it with the diversity of work produced during random times in the career. The researchers discovered a pattern across artists, scientists, and film directors with hot streaks: their work was often more diverse prior to their hot streak period than during the randomly selected times.

“The exploitation behavior during hot streaks appears consistent with several famous examples, including painter Jackson Pollock’s “drip period” (1946–1950), director Peter Jackson’s “The Lord of the Rings trilogy”, and the career of scientist John Fenn, whose hot streak arrived late in his career, but the work he produced during that period on electrospray ionization eventually won him the chemistry Nobel in 2002,” wrote the researchers. “These examples raise an intriguing question: can the exploitation behavior by itself predict career hot streaks?”

According to the researchers, “the topic that was eventually exploited is less likely to be the one explored the most recently, or the highest cited, or the most popular among the topics explored before,” which implies that “more than simply chasing after discovery through exploration, individuals appear to seek out new opportunities by deliberating over different possibilities, and then harvesting promising directions through exploitation.”

The researchers conclude that there are “identifiable regularities underlying the onset of career hot streaks, which appear to apply universally across a wide range of creative domains,” and “a major turning point for individual careers appears most closely linked with neither exploration nor exploitation behavior in isolation, but rather with the particular sequence of exploration followed by exploitation.”

Using artificial intelligence to identify patterns in large data sets, the researchers have uncovered a new understanding that may impact how talented individuals in creative domains are identified and cultivated in the future.

Copyright © 2021 Cami Rosso All rights reserved.

advertisement