Why does AI have to be nice? Researchers propose ‘Antagonistic AI’

Researchers propose the concept of Antagonistic AI, which challenges the current paradigm of overly-sanitized language models. Antagonistic AI can build resilience, provide catharsis, promote personal or collective growth, facilitate self-reflection, strengthen ideas, and foster social bonding. Techniques to implement antagonistic features into AI include opposition, personal critique, violating interaction expectations, exerting power, breaking social norms, intimidation, manipulation, and shame. It is important to ensure that antagonistic AI remains responsible and ethical by obtaining user consent, considering context, and providing framing. The dominant paradigm of AI often lacks cultural diversity and fails to reflect the full range of human values. The lack of work in the area of antagonistic AI is due to the prioritization of comfort in technology and fear among academics. However, there is growing excitement and interest in exploring antagonistic AI.