Essay.10-ZhuoyingLi-2100017764
Keywords: XAI, explainable AI, explanations
Abstract: This essay provides an analysis of explanations among humans and extends the dialogue to the domain of Explainable Artificial Intelligence (XAI). It starts by defining the essence of an explanation and the inherent human need for such discourse. Subsequently, it outlines the attributes that constitute "good" explanations. It then transitions to the realm of AI, examining the current computational methods employed to impart explanations in AI systems. Particular attention is given to algorithm-based and model-based explanations, each with its strengths and limitations in elucidating the rationale behind AI-based decisions. Finally, I suggest that a promising approach might lie in enhancing the transparency of high-performance models, such as those utilizing transformer architectures, without compromising their operational prowess.
Submission Number: 209
Loading