Reading this article about algorithmic price fixing made me check my Saved For Later list of items on Amazon yet again. I keep an eye on this list to see if Amazon is going to ever try to bait me with a the price that is exactly what I would be willing to pay. There are some categories where my returns have have been way more frequent than others - possibly signaling interest but inability to find what I like at a price that I like. Indeed, I have been notified of price drops on almost everything that sits in that list but it has never been good enough to trigger a purchase.
Just as Amazon wants to learn me and the millions of other customers they have, we have a right to engage in this kind of staring contest to learn their AI and ideally get it to do what we want. If we ceased to be paranoid and stopped engaging algorithms in ways that we can, the outcomes could be much worse. Amazon was routinely offering me rewards to deliver what I had purchased on a delayed schedule. These were not the rewards I wanted and over time, they figured out what I would accept. An engaged and informed customer can act as a wrench in the works of algorithm training and I feel that is a civic responsibility we all share today.
Recently, I had opportunity to play with a large open dataset on Kaggle with the goal of predicting if the customer would make a next transaction. The best and the brightest on the leaderboard had achieved impressive results. The outcome for the bank can only be guaranteed if their customer base operates out of a hive mind, such that the drivers of their behavior can be modeled, understood and acted upon. If people make an active effort to introduce noise in their interaction behavior, the model will not hold for long.
Just as Amazon wants to learn me and the millions of other customers they have, we have a right to engage in this kind of staring contest to learn their AI and ideally get it to do what we want. If we ceased to be paranoid and stopped engaging algorithms in ways that we can, the outcomes could be much worse. Amazon was routinely offering me rewards to deliver what I had purchased on a delayed schedule. These were not the rewards I wanted and over time, they figured out what I would accept. An engaged and informed customer can act as a wrench in the works of algorithm training and I feel that is a civic responsibility we all share today.
Recently, I had opportunity to play with a large open dataset on Kaggle with the goal of predicting if the customer would make a next transaction. The best and the brightest on the leaderboard had achieved impressive results. The outcome for the bank can only be guaranteed if their customer base operates out of a hive mind, such that the drivers of their behavior can be modeled, understood and acted upon. If people make an active effort to introduce noise in their interaction behavior, the model will not hold for long.
Comments