…and it’s not Skynet, not even remotely close.
At birth, every human is granted one gift: time. Some receive a few moments, while others have more than a century, but in the end, what we do with this time is what matters most. Did we enrich ourselves, our families, our communities, or humanity?
Many things compete for our time, families, the need to work to support them, and our hobbies that hopefully make us and our communities better. They actively seek our engagement, most with age-old techniques like love, romance, greed, etc. At the center of engagement is addiction. There is hardly a person alive who hasn’t felt the rush of scoring a goal, completing a challenging task, or checking a box on a to-do list. Our brains are hardwired to reward us when we finish something. Many of us struggle our whole lives to master our addictions. One of mine was computer gaming. Arcades in the 1980s, but I thought I overcame it with a bit of work and limiting my supply of quarters. Then PC games arrived. Sure, I stumbled around in the dark cave of Zork for a while and hung with Leisure Suit Larry for perhaps a bit too long. Still, it wasn’t until Everquest that I understood how the algorithms in these games manipulated my internal reward system to sustain my engagement. I lost a week of my life to Everquest, then one Sunday morning, I stuffed it back in its box and sold it on eBay, and I have barely touched a computer game since.
Yesterday, 60 Minutes had a story about the rise of mobile sports betting. Platforms like FanDual, DraftKings, and many more now enable gamblers to place micro-bets during a game on the next play within that game. The story explained that behind these micro-bets is an AI engine that dynamically adjusts the odds in real-time. The story then chased the obvious human side of the addiction angle. It missed the bigger story behind the story: how these AI engines are designed to leverage our addiction tendencies to increase their revenue.
The real question is, what drives the AI engines to arrive at these odds? The assumption that many users likely make is that it’s doing something similar to a parimutuel, what you often see in horse racing. This would be where the AI sees all the bets being made in real-time and adjusts the odds, taking into account the return the house requires, then pays winners from this betting pool. This is the most obvious explanation and likely the one these companies will give for how these odds are computed. Frankly, it’s the easiest for most people to understand. Unfortunately, though, this may be the furthest thing from the truth; my bet is that these AI engines are the next generation of a three-card Monte dealer; they are playing the man.
These platforms exist to enrich their owners, so at their core, it’s entirely about engagement because engagement drives revenue. Therefore, they’re optimized to keep their finger on our addiction buttons. It’s not unlike why Benny Binion in the 1950s gave the gamblers in his casino free booze; if you felt like a high roller, you’d begin acting like one. The AI engines in FanDual and the other mobile sports gambling platforms exist for one reason: to make money. Benny taught them how to improve revenue by increasing engagement. I have no proof of how these AI engines work, so the rest of my advice has no basis more reliable than my own meandering decades of software development experience…
A newbie arrives with no individual specific data set for the FanDual AI engine to operate on, so it will default to one that enforces engagement based on subtle details in their login profile. The AI will then tilt the odds in the newbie’s favor to achieve the objective of the AI, which is to set the hook in their new mark. Much like a three-card Monte dealer, who forces their mark to win one or two before sucking them in for the big bet, that they will then lose. For those who aren’t aware, a perfect three-card Monte dealer is one part illusionist, one part accountant, and one part expert psychologist. The illusionist uses extremely well-practiced sleight of hand to create the perception they require of their audience at every choice they provide, so they know which card is the “obvious winner.” The accountant tracks earnings and establishes forecasts based on various player profiles. The psychologist is constantly reading the gambler and his influencers, if there are any, and quickly determines which player profile to apply. They’ll adjust their profile assessments as things play out while always remaining focused on playing the man.
The FanDual AI engine very likely operates precisely as the highly skilled three-card Monte dealer mentioned above. It will offer up several favorable, easy-to-win micro-bets in the initial series to create some wins and begin setting the hook in the newbie. It will then shift the odds against the newbie in hopes of forcing a few minor losses to build a realistic profile of the newbie’s decision-making patterns, their timing between bets, and the degree to which they are willing to sustain risk for future reward. During this early data-gathering phase, the AI engine will likely keep the player at an overall odds advantage in an effort to make them feel like a winner.
To prove my thesis, we need only four things: a very regular, experienced FanDuel gambler and a brand new FanDuel gambler, both on different networks (one on Wi-Fi and the other on cellular, so they aren’t both coming from the same ISP network address), and sitting side by side to compare notes on the odds for matching micro-bets. I assert that these two gamblers’ odds of matching micro-bet opportunities at the same time within this platform will be inconsistent.
Perhaps today, during the Super Bowl, someone will prove me wrong, I hope you do.
AI engines will be leveraged in a great many ways, and a great many will improve our lives. A few, like those mentioned above will not. We need to recognize these for what they are, simple forms of entertainment, and not empower them to dominate our lives.