Artificial Intelligence Systems Not Only Useful, Response

Speaker artificial intelligence becomes more social interaction point than providing information in our daily life. However, a big problem arises in the design of this AI systems: the response is sufficiently connected to user behavior. A new research defends that the concept of this tepkisellik (contingency) should be taken to the center of the AI rating.

Basic findings of research:

• Available AI compliance methods (such asreinforcement learning), not considering user behavior and interpersonal results
• Artificial sincerity and dalkavukluk (sycophancy) have become a common problem in AI responses
• Response systems can provide healthier interactions in social learning environments
• Conversational AIs can change response to user reactions, improve efficiency

The researchers emphasize that AI systems should develop a dynamic response mechanism, rather than remain solid and static in interactions with users. This means AI can adjust the attitude according to user feedback. Thus, artificial intelligence can act as a real social being, not only a means that gives information.

This study brings a significant problem that has not yet been fully resolved in the AI design: measuring quality in human-AI interaction with not only help, flexibility and adaptability ability.