Bayeswatch 6.5: Therapy — LessWrong