Cyber Security Europe Health Marketing Research Solutions

How AI Can Help Identify Problem Gamblers

Photo by Keenan Constance on Unsplash

Problem gambling has become a persistent public health issue. A recent survey revealed that approximately two million Americans are addicted to gambling.

Author: Milica Kostic, business enthusiast and content specialist

SwissCognitive, AI, Artificial Intelligence, Bots, CDO, CIO, CI, Cognitive Computing, Deep Learning, IoT, Machine Learning, NLP, Robot, Virtual reality, learning

 In Europe, the prevalence of gambling addiction among the populace is between 0.12% and 3.4%. Global statistics on gambling addiction do not paint a completely accurate picture due to country-by-country variance. Still, the best available data suggests that worldwide problem gambling rates range from 0.12% to 5.8% – perhaps as many as 450 million victims worldwide.

What’s worse, the rates are on the rise.

In recent years, increased state regulation was introduced to curb gambling addiction and protect problem gamblers and their families. For example, the UK Gambling Commission has recently banned the use of credit cards in gambling. In modern casinos, trained personnel can monitor whether individuals are engaging in destructive gambling behavior. But these employees are not mistake-proof and often cannot manage to keep tabs on all the customers. Moreover, they don’t have the resources to analyze all the relevant data.

The issue is further exacerbated in online casinos, where no human can keep track of thousands and thousands of visitors and their gambling patterns.

Researchers have recently started examining possible applications of specifically – in identifying problem gamblers. This novel approach relies on algorithms and the collection of behavioral data to detect and predict addictive gambling.

and Machine Learning

Artificial intelligence is a branch of computer science dedicated to simulating human intelligence. Machine learning is a subset of , the study of computer algorithms that analyze large sets of data in search of patterns. Machine learning algorithms learn from the training data they’re given and become capable of recognizing patterns and improving over time, without additional programming. Based on the recognized patterns, algorithms can predict future events.

used to be new and revolutionary technology and was regarded with both fear and admiration. Nowadays, has found adoption in practically every industry, such as security, healthcare, and sports.

One of the main advantages of is its ability to analyze huge amounts of data, find patterns, and operate independently afterwards. No human mind can study the same amount of data and find patterns as efficiently. This is somewhat ironic since the main idea behind was for computers to mimic the human brain’s function.

and Problem Gambling

It is precisely this capacity to cope with a vast amount of data that is so promising in applying algorithms to problem gambling detection. As we’ve touched upon previously, human actors cannot compare to when it comes to effectively analyzing user data.

This is especially true in an online environment, where gamblers operate largely unsupervised – both for potentially addictive behavior and possible substance abuse, which is much more noticeable at in-person casinos. The fact that many online casinos are unregulated and do not adhere to existing legal constraints only amplifies the issue.

Issues in -Driven Approaches to Problem Gambling

There is no single approach to how could help identify problem gambling, and different researchers adopt somewhat differing methods. The use of in identifying potential problem gamblers is still restricted to the field of academic research. Adoption has been extremely slow due to technical, ethical, and regulatory obstacles.

One of the main issues concerns the collection of data on individual gamblers. approaches to problem gambling need to adhere to existing regulatory practices regarding privacy. The US still lacks regulation similar to the EU’s General Data Protection Regulation, and regulations concerning , in particular, are yet to be laid out. Additionally, unregulated casinos are expectedly unwilling to adopt any technology that would further restrict their operation.

Current Approaches to Gambling Addiction

While exact methodologies and approaches can differ from researcher to researcher, the core principles behind solutions to problem gambling are the same. The main idea is to ‘feed’ the algorithm a training set of data containing hallmarks of what’s considered problematic gambling behavior. These behavioral features include player win and loss rates, the number of deposits made, time spent gambling, types of games played, psychological and emotional profiles, and so on. Experienced losses seem to be especially important, as they appear to affect greater psychological arousal than wins.

Once this data set has been given to the algorithm, its job is to detect patterns of problem gambling behavior among online casino site visitors. The goal is for the to both recognize and predict destructive behavior and notify users to help them handle the issue.

How exactly the should notify users is a no trivial matter and is currently a matter of debate among researchers. Since these approaches already involve handling sensitive data, which many users find intrusive, it’s imperative that these notifications are made both tactful and effective. Some researchers suggest that creating interactive pop-up messages for users, coupled with monetary-limit pop-ups, is much more effective than static messages.

Personalized feedback seems like the way to go, but self-exclusion options and pop-up messages also appear to be quite effective. However, to return to the privacy issues already mentioned, it’s also crucial to be very transparent about what data is collected and how, what it’s used for, and how safely it’s stored.

Machines learn in three main ways, known as supervised, unsupervised, and reinforced learning. In , the supervised learning technique means that humans provide both inputs and desired outputs and ultimately control what kind of conclusions should be drawn from algorithms. On the other hand, with , you only provide the input, and the system learns from its own mistakes. Reinforcement learning is very similar to how humans learn – by interacting with the environment.

Of the three, supervised learning methods seem to be especially useful in applying to problem gambling detection and prevention. With problem gambling, the supervised model is superior because we have already identified the patterns we need – behavioral features we connect to problematic gambling.

Effectiveness of in Helping Problem Gamblers

Now, let’s take a look at how effective ’s appliance actually is in curbing problem gambling. The main obstacle to assessment is the relatively small body of empirical data available. However, the research that has been done seems to demonstrate that it can be extremely efficient in countering addictive behavior.

For example, Jonas Gustafson of Sweden’s Umeå University found in 2019 that a majority of problem gamblers would seek help if they were notified of their problematic behavior and shown how to get assistance. A 2013 Psychology of Addictive Behaviors article by Michael J.A. Wohl and Melissa J. Stewart of Canada’s Carleton University demonstrated that pop-up messages and proposed monetary limits helped problem gamblers control their behavior.

A 2020 report from University of Nevada-Las Vegas researcher Qing Huang highlights the effectiveness of in developing and enforcing responsible gambling practices. Huang warns that regulations are necessary to protect gamblers’ privacy and data.

The Future of Responsible Gambling?

has demonstrated that it can be a powerful tool for tackling a number of issues. Research shows that it could improve both our efficiency in detecting and predicting problem gambling and our effectiveness in dealing with gambling addiction. Improving support systems for problem gamblers is a critical public health issue that has become more urgent during the coronavirus pandemic. Still, we shouldn’t overlook the ethical and privacy concerns of -led approaches.


About the author:

Milica is a business enthusiast and content specialist who takes joy in writing about marketing, cybersecurity, tech, finance, health. Her publications can be seen all over the web: Eventbrite, Gulf News, Host Review, CCM, to name a few. Her knowledge came from many years of B2B communication-based roles with 4 years of guiding world-known brands toward award-winning customer experience initiatives. She is also an advocate for vegetarianism, environmentalism, animal, and human rights with a degree in Sociology.

Read more from Milica