Predicting Conversational Derailment

Can we spot toxicity before it happens?