Social Learning in Multi Agent Multi Armed Bandits