A new study has been published. Journal of Psychopathology and Clinical Sciences We used machine learning to analyze 74 existing self-report datasets to identify the strongest risk factors for problematic pornography use. The findings shed light on a topic that has received increasing clinical and scientific attention over the past few decades.
Pornography consumption is widespread, with studies indicating that 70-94% of adults and 42-98% of adolescents have viewed pornography in the past 20 years. While many people use pornography without problem, some users experience problematic pornography use (PPU), which is characterized by uncontrolled consumption patterns that lead to significant distress and functional impairment.
Estimates suggest that 1-38% of adults and 5-14% of adolescents may suffer from PPU. Understanding risk factors for PPU is essential for developing effective prevention and treatment strategies, especially since compulsive sexual behavior disorder, which includes PPU, is now officially recognized in the 11th edition of the International Classification of Diseases.
“PPU appears to be as prevalent as other well-known mental health problems (e.g. depression), yet it has received significantly less scientific attention until now. For example, although there is empirical evidence about risk and protective factors for PPU, our knowledge is very limited,” says study author Beáta Bőthe, assistant professor of psychology and director of the Sexuality, Technology and Addiction Laboratory (STAR Lab) at the University of Montreal.
“At the same time, theoretical models in our field suggest that several different factors contribute to the development of PPU and that these factors may interact with each other. The advent of artificial intelligence-based data analysis methods (compared to traditional statistical methods) has allowed us to consider these complex issues and include hundreds of potential risk and protective factors in our studies.”
For their study, the researchers solicited data from 98 laboratories around the world, ultimately collecting 74 datasets from 16 countries. Comprised of more than 112,000 participants, these datasets included both published and unpublished data and assessed PPU using a variety of validated scales, including the Problematic Pornography Consumption Scale, the Cyberpornography Use Inventory, and the Brief Pornography Screen.
To analyze these datasets, the researchers employed a random forest model, a machine learning technique based on classification and regression trees, which can simultaneously consider a large number of variables and their complex interactions, allowing them to reliably identify key predictors of PPU.
A random forest model was applied to each dataset separately, with the PPU score as the dependent variable and various potential predictors as independent variables. The researchers then used meta-analysis techniques to combine the results of these models to ensure that the results were generalizable and reliable.
The strongest predictor was frequency of pornography use: regular pornography viewing was found to be strongly associated with PPU, suggesting that frequent pornography users are more likely to experience problematic patterns of use.
“Individuals who frequently view pornography may be at higher risk for experiencing problems with use,” Bothe told PsyPost, “however, it is important to keep in mind that high-frequency pornography use may manifest without PPU (e.g., because of high sexual desire), and self-perceived PPU may be present even with low-frequency pornography use (e.g., because of moral condemnation of pornography use). Thus, information about a person's frequency of pornography use alone is not sufficient to determine whether their use is problematic.”
Emotion avoidance motives were also a significant predictor: those who used pornography to avoid negative emotions such as stress and anxiety were more likely to develop PPU. This finding highlights the role of pornography as a coping mechanism to manage emotional stress.
Moral conflict also emerged as a significant predictor. This refers to the conflict individuals feel when their pornography use conflicts with their personal values and moral beliefs. Individuals with higher levels of moral conflict were more likely to report PPU, indicating that internal conflict regarding pornography use may contribute to problematic patterns.
Sexual shame was also identified as a significant predictor: people who feel ashamed about their sexual behaviors, including pornography use, were more likely to develop PPU, suggesting that feelings of shame and guilt may exacerbate problematic use patterns.
Stress reduction motivation was also a significant predictor. Using pornography as a way to cope with stress was strongly associated with PPU. This finding highlights the importance of dealing with stress and developing healthier coping mechanisms to prevent the development of problematic use.
The findings indicate that “individuals who experience more negative emotions and use pornography to control them are likely to experience higher levels of PPU,” Bothe said.
Other notable predictors included the duration of pornography use at one time, fantasy-driven motivation, and feelings of guilt. General psychological factors such as anxiety and depression were also significant predictors.
Consistent with previous research, this study found that men were more likely to experience PPU compared to women, but gender, although a (statistically) significant predictor, was a relatively weak predictor.
“Based on previous findings, we expected that gender would be a significant predictor of PPU (i.e., the problem is more prevalent among men compared to women and gender diverse individuals),” Bothe explained. “However, somewhat surprisingly, gender did not emerge as a significant predictor in this study, not even making it into the top 10 predictors. This finding highlights the importance of being inclusive in pornography research and not just focusing on men's experiences if we want to better understand this phenomenon.”
However, as with all studies, there are some caveats to consider. Reliance on self-reported data can introduce biases such as recall bias, and there is an over-representation of Western, Educated, Industrialized, Affluent, and Democratic (WEIRD) countries in the data, limiting the generalizability of the findings. Future studies should strive to include more diverse populations to increase the applicability of the results.
Furthermore, studies have shown considerable heterogeneity in results, suggesting that further research is needed to fully understand the factors that contribute to PPU. Longitudinal studies that follow participants over time may provide more detailed information about how PPU develops and changes.
“We're looking to broaden the scope of our research to include more underserved or underrepresented groups,” Bothe said.
The study, “Discovering the Most Robust Predictors of Problematic Pornography Use: A Large-Scale Machine Learning Study Across 16 Countries,” was authored by Beáta Bőthe, Marie-Pier Vaillancourt-Morel, Sophie Bergeron, Zsombor Hermann, Krisztián Ivaskevics, Shane W. Kraus, Joshua B. Grubbs, and the Problematic Pornography Use Machine Learning Research Consortium.
