Analysis of visual communication design of building energy management system under the background of artificial intelligence

This paper discusses how to solve the fundamental problem of artificial intelligence (AI) being ambiguous due to multiple heterogeneous input sources in building energy management systems. This ambiguity reduces the trust people have in AI decisions and diminishes the effectiveness of energy management processes. The proposed solution has a novel neural visualization communication framework that enables the user to visualize AI decisions. There are two major breakthroughs in the study that are discussed: the development of a visualization technology for making AI decisions transparent for non-specialist users by extracting and translating effective physical meaning from attention weights, and the implementation of a user-state-driven mechanism for the interface optimization strategy. The applied long short-term memory-attention model allows for the effective retrieval of meaningful time periods and feature times that influence the energy consumption predictions that generate the “decision weights.” The proposed method will also imply the use of eye movement studies in investigating the interrelation between the features of the interface and the cognitive load of the user. The last stage of the study will involve the treatment of signals generated in response to the AI decisions by applying specialized software developed for this purpose. The use of high attention weights prompts the presence of a “pulsating border” to catch the eye; the use of low confidence intervals prompts the presence of a “translucent mask with a question mark icon” to signify danger, and high cognitive load allows the interface to be simplified without any effort. According to research findings, this slow and logical design makes it possible to enhance the decision-making ability of users by 50% and increase their trust rate from 2.26 points out of 5 points to 4.4 points.