Over the years, the gambling industry has attracted a huge amount of attention, as researchers attempt to understand how and why we make the decisions we do, whether at the roulette wheel or the slot machines.
A common seam of study is around how people choose from one of the many slot machines they're presented with when entering a casino. Such decision making is described in so called 'bandit algorithms'.
Interestingly, a team from Yale believe that such algorithms could also be valuable in tackling HIV. The work, which was documented in a recently published paper, aims to improve our ability to detect 'hotspots' of HIV infection that can help public health teams identify areas of potential infection among people that don't know they carry the virus.
“Different methods have been proposed to identify hotspots, but there is no consensus on where to look,” the authors say. “Even setting up where common sense dictates—like places with high HIV prevalence or concentrations of people at risk of infection—doesn’t guarantee a high yield of new diagnoses.”
Spotting the hotspots
The researchers wanted to try something new, and turned to a bandit algorithm called the Thompson Sampling (TS) for inspiration. When gamblers use the TS, they tend to make their bets based upon the best current information. Each successful (or unsuccessful) game adds to their library of information for that machine.
The researchers deployed this logic to AIDS programs, and suggests medical teams can select hotspots based upon the latest information on an area, and therefore the likelihood of success.
They tested the algorithm out on a number of locations that had been targeted for HIV testing. Data on each location is updated whenever testing occurs. Their computer simulation pitted the TS driven approach against a number of other strategies for deploying the testers most effectively. Amazingly, the TS strategy seemed to trump all others.
“If you were using Thompson Sampling in a casino, a player wouldn’t make a large bet on one machine, they would play for a while to hone in on the best machine to play their quarters,” the authors say.
Whilst it seems far fetched to apply something from gambling to healthcare, the bandit method has already been applied successfully in other fields, whether it's hunting for downed airplanes or searching for the best place to drill for oil. This is the first time it's been used in the screening for infectious disease however.
The researchers will next seek to test their algorithm in the real world, with the team hoping that a successful implementation will enable a more effective delivery of AIDS programs. With estimates suggesting that 14% of existing HIV bearers are undiagnosed, it's an improvement that can't come quickly enough.