A BBC story reveals the secret to winning at rock-paper-scissors. The answer is something that we dog trainers tell our human clients all the time: if something works, an animal is going to do it again.
This means that if a dog hears something in the hall (the neighbors) and barks (the behavior), and it WORKS (the neighbors go away), then the behavior has been reinforced and will most likely occur again in the future. The dog doesn't know the people went into the elevator. As far as he is concerned, his vocalization has caused them to evaporate. I think dogs think pretty highly of their powers.
Ditto if your dog gets a whole turkey off of your kitchen counter. The behavior of counter surfing has just been bolded, italicized, underlined and capitalized. Jumping on the kitchen counter is an awesome idea!
In humans, we see this happening all the time. The behavior doesn't even have to be rewarded every single time. Or at all. Think about how many times Ralph Nader has run for president. A variable rate of reinforcement can be a powerful thing. I'm thinking about a time at a bar a few weeks ago when a guy asked me, "Great legs, what time do they open?" Guessing this had worked for him at least once in the past.
It turns out that the same thing occurs when humans play rock-paper-scissors. According to the BBC:
Scientists recruited 360 students and divided them into groups of six. Each competitor played 300 rounds of rock-paper-scissors against other members of their group. As an incentive, the winners were paid - in proportion to their number of victories. To play smart, classical game theory suggests players should completely randomise their choices - to remain unpredictable and not be anticipated by opponents. This pattern - where both players select rock, paper or scissors with equal probability in each round - is known as the Nash equilibrium...However, on closer inspection, the organisers noticed a surprising pattern of behaviour. When players won a round, they tended to repeat their winning rock, paper or scissors more often than would be expected at random (one in three).
Therefore, the solution, is not to stick with what works. But let your opponent fall into that trip. Just like a dog will very likely revert to a "sit" if that's the last thing he was rewarded for, you can expect your opponent to lean on their most rewarded move.
Don't understand? Don't worry about it. It's just a silly human game. Your dog still thinks your a genius.