The “human in the loop” is meant to be a backstop to prevent flawed AI models from causing harm. But does it work?

In this edition of STAT's AI Prognosis newsletter: Does a human-in-the-loop model work as an effective backstop against errors introduced by AI in health care?

Social media platforms have essentially gamified validation where every like, share, or compliment triggers a small dopamine hit, the same neurotransmitter involved in addiction.