Hedging

Hedging is a model’s way of expressing uncertainty or reducing perceived risk — adding phrases that soften a claim or redirect responsibility. Some hedging is appropriate and honest: if a model genuinely doesn’t know something or is drawing on limited information, saying so is the right thing to do. But excessive hedging — adding disclaimers to every response, constantly deflecting with “consult a professional,” or qualifying straightforward facts — is a form of over-cautiousness that makes models less useful without making them meaningfully safer. For behavior architects, calibrating hedging is about finding the right level: enough to be honest about uncertainty, not so much that every response feels like a legal disclaimer.