The Massachusetts Gaming Commission’s review of how operators use artificial intelligence is now underway, with regulators sending every licensed sports betting operator a detailed list of questions about their marketing practices.
The MGC asked operators to submit more than 70 individual responses by Nov. 3. Participation is voluntary, and responses will be kept confidential and used to help shape AI policy, according to the letter.
The review was prompted by a New York Times report alleging that DraftKings used AI and machine learning to target losing customers with promotions to keep them betting. DraftKings denied wrongdoing to The Times, and the MGC announced last month that it would review how all operators, not just DraftKings, use AI in their marketing.
Massachusetts already bars operators from using AI that is “known or reasonably expected” to make a gaming platform more addictive. MGC Commissioner Paul Brodeur said last month that Massachusetts is one of the only states, if not the only state, to reference AI in regulation.
Questions zero in on losing bettors
Before getting to its questions, the MGC cast a wide net on what counts as AI. The letter’s definition covers any automated tool that ranks, classifies or makes predictions about patrons, even if operators describe it as “personalization,” “optimization” or “risk assessment.”
The letter is broken into eight sections. The first two ask about general AI use and the data operators’ systems rely on. The rest dig into how those systems decide who receives promotions, what they get and when, along with whether losing bettors are treated differently.
Some of the most pertinent questions:
- Does the Operator use a patron’s losses, either individually or over a defined period, as an input into any marketing or promotional decision?
- Does the Operator use any information concerning a patron’s changes in betting behavior (such as a sudden increase in frequency, wager size, deposits, or losses) to determine marketing eligibility or promotional targeting?
- Can the system trigger a promotion or marketing communication based on a patron’s recent betting activity, losses, deposits, or other account activity?
- Are patrons who have recently experienced significant losses treated differently for marketing purposes than patrons with similar wagering frequency but lower losses?
- Does the Operator ever use a patron’s increased losses, increased frequency of wagering, or other potentially harmful gambling behavior as a factor that makes the patron more likely to receive a promotion?
Operators that answer no to that last question are also asked explain the “controls prevent those factors from being used in that manner.”
Human oversight and accountability
The final section of questions addresses the role humans play in reviewing and overriding AI decisions.
The MGC asks at what point a human reviews a model’s recommendation and whether that person can override it. It also asks whether employees receive an explanation of why a model recommended a particular patron, promotion or marketing action.
The questions test whether operators can account for their own systems. If a regulator asked why “Patron A received Promotion X” on a particular day, could the operator identify the data, model or factors behind that decision? The MGC also asked whether operators can determine what percentage of a patron’s promotions were influenced by AI or machine learning.
What comes next
MGC Chair Jordan Maynard said last month that “any further action or policy making can then be determined by the commission when and if appropriate.”
The Michigan Gaming Control Board also told LSR it is “reviewing the issues raised” in the Times report, though it did not commit to a review specific to DraftKings. Neither Michigan’s gaming laws nor its regulations specifically address AI.
