MEA System Automates Description of Complex AI Models

Artificial intelligence systems are increasingly used in critical areas such as medicine, finance and law. But we are forced to explain how these models have decided. The new research shared in ArXiv proposed a system called **MEA (Multi-Agent Explanation Agent). This system will make it easier to understand the forecasts of complex models.

The main problems associated with existing explanation methods:

• Very difficult to understand multi-dimensional outputs for domain experts
• Expert knowledge required to choose the best description
• Time to combine evidence from different tools
• There is a lack of transparency in high risk decision areas

MEA is comprised of multiple artificial agents that use the**l mechanism. Each agent is called different description strategies and successful ones receive more rewards. In this way, the system can automatically select the most appropriate descriptions. With the reduction of human intervention, domain experts understand the models better without technical knowledge.

This approach can improve the trust of AI applications. In particular, in the health and finance sector, it is legal and ethically critical to understand the logic of the model before deciding. Systems like MEA can be an important step in solving **AI’s closed box problem. You can access the full details of the research from arXiv.