When some people think of artificial intelligence (AI) and gaming, images of players trying to cheat the system or operators rigging play may come to mind. But AI could become one of the most valuable tools to date for responsible gaming (RG).
Consider this: All of the data now available to operators about time and type of play, amount wagered, player behavior and more can also be used to trigger RG cues.
When players use a loyalty card for in-person gaming, that information is part of a warehouse of data a casino has on hand. Similarly, cookies used in tracking online play also gather information from a player. This information can be sorted using AI to determine individual play patterns and behaviors. While great as a marketing tool for targeted campaigns and rewards programs, that information can also be used for harm mitigation as well.
AI Can Identify When Play Becomes Risky
Former Colorado gambling regulatory chief Dan Hartman is now an associate partner at GMA Consulting, a gaming, hospitality, integrated resort, and sports & entertainment research and consulting firm. Earlier this month at the Global Gaming Expo (G2E) in Las Vegas, Hartman said that deploying AI is becoming more accepted on the regulatory front, particularly when it comes to RG.
“For the responsible player and the gambling industry, it can be a tool to ID people where play isn’t good for them,” he said. “We can find a way to delay them before they hit the button to play. We can see they start making choices that aren’t smart and we can send them information to help them make smart choices.”
Francesco Rodano, chief policy officer at Playtech, noted that pattern matching, where operators can look at the previous 15 days of play, can be used to see if there are any statistical variances for players. Rodano said Playtech looks at 70 behavioral features like amount bet, sites used, frequency of play, deposits per week and more to determine risk levels.
Using random forest, a machine learning algorithm that combines multiple decision trees to get to a single result, operators can determine risk levels of players based on low-, medium- and high-risk play. A decision tree starts with a single question – Should I play? – and then asks a series of follow-up questions to determine the next action. These could be “Do I have money to use?” or “Do I have time to play?” Depending on the answer to each question, the player is funneled to the next branch of the decision process.
Personal and Digital Outreach
According to IBM, “decision trees seek to find the best split to subset the data, and they are typically trained through the Classification and Regression Tree (CART) algorithm.”
Using this process, operators can track the interactions with a player to see if they move into a high-risk category by increasing game-play time, spending more money or changing betting patterns. AI can parse this data, Rodano said. Land-based casinos can sort levels of risk through unique IDs associated with a player’s loyalty card or login. If a player’s risk level rises, a customer service agent can call a player whose activity has moved into a more high-risk level of play. Because the customer service agent would be able to see changed patterns of play via a dashboard AI has compiled, this helps flag potential problems earlier.
Further, online gambling and sports betting platforms with AI engaged can utilize pop-up messages to make suggestions to players to take a break or set a budget if it recognizes excessive or behavioral changes. The chat bot can also launch a conversation with a player to see what has changed in style of play.
“If the operator can intervene at the right time, they find it can mitigate harm,” Rodano said. By deploying chat bots that utilize large language models, the AI can have a “conversation” that is unique to the player. This personalization can be 20 times more effective than just imposing broad betting limits, Rodano said. Using predictive AI, operators can help players make their own choices to limit their play and mitigate losses, which, in turn, he said, can keep a customer loyal by having their best interests at heart.
Using AI to Train Customer Service Agents
This AI could also be used to help customer service agents in their outreach to high-risk players. By offering suggested scripts based on the data collected as well as sentient analysis, a more effective outreach can potentially mitigate harm.
Sentiment analysis, Rodano said, would read the mood of the player through the interactions in the chat, learning for instance, if a player showed a lack of control or anxiety based on keywords used in the chat.
This use of a large language model helps the bot learn and become better with each use. This doesn’t supplant traditional RG tools, but rather supplements them and can be the first step to intervention.
One barrier that will be hard to overcome, both Rodano and Hartman said, is sharing that mountain of data that each operator collects to use it to help players. Knowing that a high-risk player has multiple accounts at both brick-and-mortar casinos as well as digital can trigger safety measures more easily. But sharing information willingly is key.
This data sharing, which is easier to accomplish in the United States than in Europe because of privacy laws, needs to be embraced by all stakeholders to do everything to minimize harm to gamblers, said Brianne Doura-Schawohl, founder of Doura-Schawohl Consulting LLC, a global government relations firm that specializes in responsible gaming policy. She likened this embrace of shared data to putting your seatbelt on every time you get into a moving vehicle–you don’t anticipate an accident, but you are happy when you have it on should you get into one.
No one anticipates having a problem gambling, but using the technology available and sharing it across platforms and operators, can lead to a safer experience for everyone.
That, combined with thoughtful human interaction after initial AI contact can make a safer environment for all players.
