Despite the chant that correlation is not causation, some researchers believe the design of scatter plots nudges us to the wrong conclusions. Can a change in their design lessen that risk?