There are enough sensors in a car to know when a collision has occurred. The car is capable of capturing touchscreen use metrics. Bonus points if you can find metrics for the same region/timespan for cars with tactile controls. You can then do any year 1 statistics algorithm to see if using the touchscreen is correlated with collisions. Yes, this doesn't prove causation. A correlation is enough to raise the flag about a safety issue.

EDIT: A decent statistical analysis will also uncover if using the touch screen is negatively correlated with accidents -- if people that use maps in general are less likely to get into an accident. This again won't prove causation but would be of interest to the general public and regulators.