Alternating between “Algorithm Madness” and “Caged by Algorithms” seems to be the new norm. There are strategies worth considering to avoid the game.
Recently I was told by an insurance broker would I accept less coverage on a pickup I had purchased, as their system’s underwriting AI (Artificial Intelligence) “thinks” I paid too much for the truck.
Then in an semi-spam email I picked this out about “auto-pricing”:
Most bullion websites run on an algorithm that automatically prices products based on demand. If Goldbacks or other product appear on zerohedge or other publications and there is a large uptick in traffic, these bullion websites will automatically raise their prices to earn more money on the increased sales volume. It is best to buy when no one else is looking for your specific product to get the best price.
Then in Amazon Prime and its algorithm pricing models featured in several articles I skimmed while in isolation.
In each case a significant impact of the output from an algorithm is affecting us as the end consumer. An Underwriting Algorithm that doesn’t have direct access to the market values, a Pricing Algorithm that will punish those who follow the popularity side of the market, and a combined Shipping/Price Algorithm that works against the value proposition the consumer was sold.
So how to mitigate these sorts of impacts?
One suggestion is so simple – Do Not Play the Game!
Specialty/Collectors Car insurance on non-exotic cars seems to run between 0.5% to 2.5% of “agreed value.” (This sort of ratio seems to apply to radio collections as well.) The AI being used in my case expected a contemporary vehicle not a nearly 60 year old trophy winning classic. In many cases your big name insurance company assigns-out collector’s policies but retains the administration inhouse. Misuse of inappropriate AI is not uncommon. I’ve asked my broker to get their central office to have a human review things before I place the coverage elsewhere. I already have one vehicle insured this way because their AI wouldn’t accept the vehicle as something that exists, and if I am shoved across on this second one, I may as well price all my vehicles with both carriers.
As I don’t buy bullion and have long observed that the premium above spot price goes up the more PR a base metal receives, it is easy to sidestep this gotcha completely. But the same pricing game applies in other markets, so there I need to be careful that my interest doesn’t drive up my potential purchase price.
The Prime game deserves a whole post on its own, as there are several different axis that as a consumer that I am hoping to optimize. Purchase Cost (all up) is one, Availability, Convenience, Time between order and receipt, are a few others, and minimization of time lost to shopping another axis. This is a complex area to evaluate, one that takes on a wholly different complexion if you ascribe a value to your own time. In my case I use Prime a lot, and largely ignore the cost impact on small sundries, but do price check is an item exceeds $100 or experience has suggested that Prime is a poor value proposition for that particular product class. Not a perfect work-around, but I value my time enough that losing more to questing for possible but not always available bargains quickly has a poor ROI.
So you can work the algorithms, but not head-on disputing their actual calculation process, but by seeking human pricing, avoiding purchasing against algorithms, or setting up your own rules & heuristics to mitigate their impact.
73
Steve
K9ZW
With Varying Frequency - Amateur Radio Ponderings 