Zona de Azar USA – G2E: The Power of AI and Machine Learning in Detecting Risky Gambling Behavior
USA.- October 19th 2023 www.zonadeazar.com By leveraging advances in artificial intelligence and machine learning to create predictive algorithms for responsible gambling, casino operators can reduce harm to players by configuring play interruptions when risky behavior is detected, according to experts in AI.
The Global Gaming Expo explored the topic last week with a session on potential technology enhancing the safety and integrity of play for users.
One way it can be done is by creating friction in the form of app-based texts, emails, and push notifications to the player or as self-service reminders to players regarding safer playing options.
Mike Reaves, the head of Worldwide Solutions Architecture for Betting & Gaming at Amazon Web Services, said they’re using machine learning to detect problem behavior in gaming and “trying to be a force for good.”
Reaves said they’re currently working on two systems in the betting and gaming space to help suppliers, operators, and regulators. The first is a predictive algorithm that looks at different indicators, including financial account information and wager data.
“We can create a machine-learning model using the operator’s data to try to detect when gambling behavior may become problematic,” Reaves said. “When it is, the cool thing that can be done with technology these days is you can make a notification in real time to prevent harm from being done at the time. In the old days, you got a report and saw that so-and-so lost $10,000 and you couldn’t really do much about it, other than call them up and see if they were all right and offer them a credit.”
Reaves said AWS is also working on personalization solutions using AI and machine learning like people might see on Amazon Prime Video or the Amazon e-commerce site where people get suggestions on what to purchase.
“That same sort of technology can be used in betting and gambling to maybe offer a bet someone is interested in,” Reaves said. “There’s a fine balance between giving someone a recommendation and trying to prevent problem gaming, but we’re trying to apply machine learning to all of these sorts of problems and identify solutions that are useful.”
Paula Murphy, business development manager at Mindway AI, said machine learning is a subset of AI and what they do at Mindway is teach the algorithm to replicate human decision making.
“For something like problem gambling, we look at every 10 minutes of casino play, at the sportsbook, and hands of poker and glean behavioral patterns that look at some of the same markers,” Murphy said. “Because we use expert human psychologists, they can bring a contextual analysis that you couldn’t get if you were looking at markers. We’re tracking some seven-and-a-half-million players on an ongoing basis for the operators that we work with.”
Madeleine Want, vice president of data at Fanatics Sportsbook, said the difficulty with predicting problem gambling is that it’s a “data problem.” It starts with someone who tells you what to look for by identifying confirmed problem gamblers from the past.
“You build the tools and algorithms to run that through your own customer base,” Want said. “We ask. who’s behaving similarly to those who we might not have seen? What are the correlating factors that our responsible-gaming team didn’t proactively suggest, because they weren’t aware of them? However, the algorithm has noticed them and could bring forward behavior of other customers who have fallen through the cracks. We’re brand new in this space and live in five states with many more to come. One of the things that machine learning needs is a lot of data and when you haven’t been around long and are only in a small subset of states, you don’t have enough data to train a very hungry machine-learning model.”
Want said the approach they’ve been taking in partnership with AWS is to build out the framework of how to enter the data, so it can be used for such a model. Once they have a sufficient amount of data, they’ll switch that out for a machine-learning approach.
“Another reason this is such a great data problem is that we take an approach to translate human intuition into rules and tell the system how to behave,” Want said. “That’s all contributing to a body of data that will then be used to train and score a future machine-learning approach. The data is one small component. It’s what you do with that information once you have it where the change happens.”
Becky Harris, former chair of the Nevada Gaming Control Board and Distinguished Fellow in Gaming and Leadership at the International Gaming Institute at the University of Nevada Las Vegas, said one of the challenges is that these applications are done on a jurisdiction-by-jurisdiction basis and data selection on an operator-by-operator basis that “vary dramatically.” Until the industry gets more comfortable in using AI, gaming regulators will be hesitant to rely on it, she said.
“Like everything else in responsible and problem gambling, more tools and diverse tools are beneficial and it’s good to be able to identify people acting in particular ways and help inform us.”
Harris said the lawyer in her, however, has questions about people’s civil rights and what regulators accept in terms of players being shut out of activities in which they want to be engaged.
“This is great in a mobile setting. But so much of our casino industry is land-based, so where’s the application for that?” Harris said. “I can see tracking through players’ cards and maybe this technology gets to a point where there are some real-time opportunities in AI to identify people engaging in harmful behavior. AI by itself isn’t the answer. We’ll have to look at a lot of different policy levers. The conversation with problem gambling shouldn’t begin and end with AI. We should see where it fits in on the continuum and how much confidence we have in it.”
Edited by: @MaiaDigital www.zonadeazar.com
