More witnesses should mean MORE help, right? Wrong! Darley & Latané's shocking 1968 discovery: the MORE people who witness an emergency, the LESS likely anyone is to help. Watch the simulation—see how diffusion of responsibility makes everyone assume "someone else will do it."
Participants in a "discussion" heard a confederate have a seizure over intercom. Alone: 85% helped within 60 seconds. With 4 others: Only 31% helped within 60 seconds!
Participants filled out questionnaires as smoke poured into the room. Alone: 75% reported within 2 minutes. With 2 passive confederates: Only 10% reported!
A woman screamed and fell in the next room. Alone: 70% offered help. With a stranger: 40% helped. With a passive confederate: Only 7% helped!
If each bystander has probability p of helping independently, the probability that SOMEONE helps with n bystanders is:
P(help) = 1 - (1-p)n
This SHOULD increase with more bystanders! But the bystander effect means each individual's p DECREASES with more people. Empirically:
p(n) ≈ p₀ / √n
This "diffusion penalty" can outweigh the benefit of more potential helpers, creating a U-shaped or even declining curve!