>Also if I go drinking with someone I don't know well, the risk of their becoming violent is quite high
In what kind of society or social circles is this true? Here it's common to socialize and meet strangers over a drink or two, and the risks of their becoming violent is near zero.
Many drugs promote socialization. To the point that you have to question the distinction between drugs and food/drink. Take sugar. It's often the focus of social gatherings. But the bad effects of sugar arguably outweigh those of alcohol, given the epidemics of diabetes and obesity.
Alcohol happens to have 10,000 years of cultural tradition behind it. Plus, it pairs well with food.
Pick a random person at a random moment in time. What is the probability that the person is under the influence of alcohol? No more than 3% is my guess, but some of my debate opponents might disagree, so let's appease them and say 6%. Keeping on randomly choosing (person, time) pairs, taking note of whether the person you pick is engaging in any violence. Ignore any violence done by anyone other then the person you randomly picked. 40% of the violence you note occurs during the 6% of the time the person you picked is under the influence of alcohol.
Like someone here writes, correlation does not imply causation, but correlation is evidence of causation, and that is a strong correlation. Specifically, if the person you picked is under the influence, he or she is 10.4 times more likely to be engaging in an act of violence than if he or she is not.
In what kind of society or social circles is this true? Here it's common to socialize and meet strangers over a drink or two, and the risks of their becoming violent is near zero.