A pioneering AI system successfully identifies violations of social norms. Utilizing GPT-3, zero-shot text classification, and automatic rule discovery, the system categorizes social emotions into ten main types.
Researchers at Ben-Gurion University of the Negev have successfully developed an AI system capable of identifying social norm violations. This pioneering project, supported by DARPA funding, represents a significant step towards automating the detection of social norm breaches. While diverse social norms are prevalent worldwide, this system efficiently categorizes violations into a few general groups.
The cutting-edge AI system was meticulously crafted by Prof. Yair Neuman and his collaborator, engineer Yochai Cohen. Utilizing advanced techniques such as GPT-3, zero-shot text classification, and automatic rule discovery, the system employed a binary framework consisting of ten social emotions as distinct categories.
The motivation behind this research stems from the Computational Cultural Understanding (CCU) program, which DARPA commissioned to develop cross-cultural language understanding technologies. The ultimate goal is to enhance situational awareness and improve interactional effectiveness for Department of Defense operators. Notably, miscommunication across different cultures not only disrupts negotiations but also poses a potential risk of escalating conflicts, as highlighted in DARPA’s rationale for the program.
Their remarkable findings have been recently published in the esteemed journal Scientific Reports. Prof. Neuman and his team successfully trained the AI system to recognize ten key social emotions: competence, politeness, trust, discipline, caring, agreeableness, success, conformity, decency, and loyalty. The system effectively characterizes written situations under these ten classifiers while discerning whether they exhibit positive or negative aspects.
Prof. Yair Neuman, the head of The Functor Lab in the Department of Cognitive and Brain Sciences at BGU, reflects on the significance of their preliminary work, stating, “This study provides compelling evidence that our approach is accurate and has the potential to accommodate an even broader range of social norms.”
The innovative AI system developed by Prof. Neuman and his team represents a promising advancement in the automatic identification of social norm violations. Its potential implications for promoting cross-cultural understanding and mitigating conflicts hold great promise for the future.