Twitter/FriendFeed updates announced Amazon.com was down for at least an hour this morning. It has since come up and down a few times. CNet, Mashable, TechCrunch and, thus TechMeme, eventually realized it too. It’s up for me now.
I have no idea what happened, but I know Amazon gets a surge of buzz/traffic when it returns. I’ve seen Twitter leverage this strategy masterfully, reaping the rewards of variable scheduling to maximize conditioned behavior.
Wikipedia’s description of operant conditioning has this to say about fixed and variable scheduling of stimulus:
“According to the laws of operant conditioning, any behavior that is consistently rewarded, every single time, will extinguish at a faster rate while intermittently reinforcing behavior leads to more stable rates of behavior that are relatively more resistant to extinction. Thus, in detection dogs, any correct behavior of indicating a “find,” must always be rewarded with a tug toy or a ball throw early on for initial acquisition of the behavior. Thereafter, fading procedures, in which the rate of reinforcement is “thinned” (not every response is reinforced) are introduced, switching the dog to an intermittent schedule of reinforcement, which is more resistant to instances of non-reinforcement.”
Image above from Reinforcement Schedules (VR line shows maximum impact from Variable ratio schedules)
Applying this to Twitter’s intermittent downtime and you can see how Twitter awareness is reinforced every time Twitter comes back from an outage. Services like Twitter need people addicted and, ironically, random outages can help drive the addiction. Hopefully, Amazon’s lost sales make it too painful for them to follow a similar approach…