Even without explicit collusion, pricing algorithms converge on price-fixing strategies

Originally published at: https://boingboing.net/2019/02/14/skynet-price-gouging.html


Hypothetically, then, it should be possible to fight pricing algorithms with purchasing algorithms. If it’s based on trial-and-error pricing, a purchasing algorithm with enough buyers behind it will move the market pricing consistently down until the pricing algorithms hit their basement threshold setting. Right?


Is this because the relationships are generally mis-perceived. In a one-on-one model, the purchaser and the seller are generally in direct competition, with each trying to maximise their own value from the transaction.

When you add a second seller they’re aren’t in competition with each other, they’re both in competition with the purchaser. In some repcts this makes the sellers ‘friends’, in an enemy-of-my-enemy way.


Wait, so… Wait…

So… I don’t get it. The bots are going to make everything sell for a penny and also $10,000, and this is unfair? Wait, is it Russian collusion? Asking for a friend.

Can someone explain to me why the price that the bots found isn’t the Nash equilibrium? I understood that the ‘optimal’ price is where both agents could maximise their profits, but the paper suggests that this is lower than where the bots settled.

Are the bots both making less profit than they could? If so, why?

1 Like

Likely because some players don’t know the actual equilibrium strategies of the other players. Also, because some players are not rational and/or makes error and don’t actually make the best decisions.

Bad AI!
Bad AI!!
(fetches spray bottle)

1 Like

This topic was automatically closed after 5 days. New replies are no longer allowed.