Bayesian supermarket

Earlier today, Marta and I tried an exciting experiment while shopping for food. Before we got into the supermarket, we proposed that each of us (including Kobi; 6 years old) guessed how much we’d spend and then, after the first aisle, we’d look at how much we were currently spending (using the time-saver device, which displays real-time data) and revise our guess.

He didn’t really get the updating part — he started with a guess of $\mu=$ £20, which, after seeing data $y_1 =$ £30 at the end of the first aisle (less than half-way through), he revised to $\mu\mid y_1=$ £25… When we explained that he had to go higher because we were already spending £30, he misunderstood and thought we should simply spend more thus suggesting we bought every single crappy item (from chips to biscuits), matching each with one of his friends who would normally have them for snack.

For the record, all the prior guesses were way off the final outcome — but in all fairness, this was mostly due to a massive outlier (1kg of coffee beans, which we hadn’t considered in our list, but was on offer — we don’t drink that much coffee…).

comments powered by Disqus