This
Atlantic piece on the history and science (or lack thereof) behind Alcoholics Anonymous (AA) is interesting a many levels, including in particular how a treatment with very little in the way of real evidence behind it ends up becoming so thoroughly enmeshed in government policy.
Similarly interesting are two critiques. One is
from Scientific American's blog. It is incorrect that AA is free; while it has no explicit money price, donations are solicited at meetings and, of course, the time cost is very large for those who take it seriously. Still, the general point of comparing costs and benefits is a good one, and suggests heterogeneous optimal treatment choice depending on factors such as value of time. This critique also offers no comfort for the (common) practice of sentencing people to participate in AA. Changing the Atlantic's conclusion from "no good" to "who knows?" does not justify compulsion.
The other critique is
from New York magazine. I wish this one had more detail; it sounds like researchers are using what I would call a randomized encouragement design by randomly assigning people to formal preparation for the 12-step programs. This design, of course, estimates a local average treatment effect, which is then confused in the write-up with the average treatment effect on the treated. Or so I suspect. Also suspect is the apparent mono-focus on abstinence as an outcome. Real success, it seems to me, is leaving people able to drink in moderation, not condemning them to never drink again (not that there is anything wrong with that ...).