When New York Congressman Paul Tonko unveiled plans to introduce new legislation seeking federal regulation of the legal sports betting industry in the United States, one of the most talked-about wrinkles in the bill's outline was one seeking to ban the use of artificial intelligence to track the gambling habits of individual players.
Tonko's intent here was clear: In the next bullet point, he advocated for prohibiting “the use of AI to create individualized offers and promotions to customers.” But the legislative outline made no mention of the potential collateral damage of such a ban – specifically, the hinderance of emerging AI technology that's capable of identifying and preventing problem gambling among individuals, as well as flagging potential integrity breaches.
“In order to be able to really scale the actions you want to take, AI can do things that humans can't,” Jonathan Inbar, a marketing data scientist and director of strategic services at Optimove, told Sports Handle. “It can analyze patterns in order to flag out, quickly, any irregularities or potential harmful behavior. I’ve been in discussions with operators who’ve told me they’re trying to look at a random sample of customers and see if they behave in a certain way that should wave a red flag, but they can’t do it by themselves in analyzing millions of customers.”
“Of course AI is being used for commercial purposes, but a lot of operators are using AI for player protections and consumer protections and doing great things with it,” added Kasra Ghaharian, a postdoctoral research fellow at UNLV's International Gaming Institute.
Matthew Holt is the president and founder of U.S. Integrity, a global leader in detecting betting irregularities in sporting events. U.S. Integrity collects real-time betting data from 47 sportsbook operators, with Holt calling AI “a tool to help sift through that top level of abnormality.”
“It's becoming more and more important,” he added. “A human being can only investigate so much in a real-time basis. We’re constantly working on AI and how we can use it to better identify and sort through those top-level abnormalities. It needs state legislators and regulators to enact strong data mandates and then enforce those regulations so integrity monitors are receiving all the data they need to detect abnormalities.”
Called the SAFE (Supporting Affordability and Fairness With Every) Bet Act, Tonko's legislation has yet to be filed and remains a work in progress.
Sign Up For The Sports Handle Newsletter!
To this end, when Sports Handle contacted his office to ask whether AI would be unable to be used to track suspicious or problematic behavior among individual players, a spokesperson replied, “The legislative intent of this framework is to stop the predatory practices of individualized promotions. By no means does it seek to limit any tools that are currently being used to identify and safeguard against risky behaviors. While there have been instances of European countries employing AI to identify problem gambling, we have yet to see this practice meaningfully or effectively applied anywhere in the United States. Regardless, we're confident that the final legislative text could address this balance issue in a way that prioritizes the safety of consumers.”
Gambling regulators in New Jersey, a state which requires operators to use technology to identify and address at-risk players, might have something to say about whether customer data is being used to effectively combat addictive behavior in the U.S., but there at least appears to be some legislative wiggle room.
States vs. feds and unintended consequences
Tonko's first effort to bring federal oversight to sports betting was unsuccessful, which could have had something to do with the 2018 Supreme Court decision that cleared the way for sports wagering to occur outside of Nevada.
“When the Supreme Court made its decision, I thought that the federal government would detach itself from the issues and let it be governed at the state level,” said Vince Gennaro, a clinical associate professor at New York University's Tisch School for Global Sport. “So, in some ways, given all the other challenges of federal government, it’s interesting that they’ve chosen to dive into this topic, one that has already gotten a Supreme Court decision saying it’s a states’ issue.”
“Broadly, if we got rid of the Wire Act and didn’t have to pay the handle tax, we’d be happy for the federal government to take more of a role,” said Lance Agostino, director of B2B Sportsbook at Betsson. “But the way Congress moves, I don’t think anything will happen anytime soon.”
In light of ongoing betting investigations involving the likes of Los Angeles Dodgers slugger Shohei Ohtani, Toronto Raptors center Jontay Porter, and the Temple men's basketball team, Gennaro said, “It would be surprising if either the leagues or the betting organizations themselves weren't all over this issue – the integrity of the sport.
Representative Tonko is coming at it from the standpoint of the mental health of the betting public and the addiction issue. I know I’ve seen ads where they’ve said, ‘Here’s our betting service, and if you have a problem, here’s the number you call.’ I guess maybe one value the federal government could add is to make suggestions to the industry as to what disclosures they’d like to see.”
When asked if he could see the value in limiting the use of AI to tailor promotions to individual bettors, Gennaro replied, “Could that accelerate someone’s involvement? I suppose it could, but we’re talking about something at the margins here. The cat is out of the bag with sports betting, so I think the kind of things we’re talking about are at the margin and aren’t necessarily high-impact solutions to problems. It’s there in front of us if we’re so inclined to bet, I don’t know that a series of sports betting ads to under control bettors is going to make them an out of control bettor.”
For his part, Holt had this to say on the issue: “Using AI, these ads are going to be more personable and more relevant, but encouraging people to wager in a regulated market, I always think there are going to be some positive effects — at least aiming them toward the regulated market where there’s addiction assistance that people can get. If we don’t help the efforts of the regulated operators, then we’re sending people into the illegal market.”
Identifying high-risk players
In more mature foreign markets, the use of artificial intelligence in providing player and integrity protections is old hat, said Ghaharian, who went on to wonder just how, exactly, Tonko's legislation would define AI.
“The industry has done player segmentation through cluster analysis for years and years and years,” Ghaharian pointed out. “Cluster analysis is technically machine-learning, so is that AI? AI is a broad definition, so they would have to define what they mean by AI.”
Back before responsible gambling became what Optimove's Inbar described as “the hot buzzword in the European gambling market” around the year 2019, he recalled, “When we started researching a solution, the most common approach was sending questionnaires or surveys to customers [saying], ‘Tell us about your gaming behavior.’ We don’t need to be geniuses to understand that the bettors who are getting those questions will answer how they want to answer, not what they actually do. There are all sorts of self-serve options, like, 'Click here to self-exclude yourself.' These were all very reactive approaches. There wasn’t really a good, proactive solution to raise alerts.”
Since then, Optimove has developed a predictive model that assigns each individual gambler a risk score based on their behavior on various betting apps. For example, Inbar said, “You take everyone that has a score of eight or more and you train a machine-learning algorithm in how they behave. From there, you can arrive at a probability for high-risk players.”
“Very irresponsible bettors are very similar to VIPs,” Inbar continued. “What you don’t want is to treat them the same way with your marketing promotions, to throw more gas into the fire. At the same time, you want your true VIP players to get the treatment they deserve, and using these type of models helps distinguish between the two.
“It’s not just about having an automated model. We need real-time alerts to flag irregular behavior. If someone has deposited more than X amount of money or is logged in for more than X amount of time, there needs to be an automated intervention.”