Detailed Review of Explainability Needs in Generative Artificial Intelligence: Diverse Applications, Current Insights and Future Opportunities

Artificial Intelligence (AI) is emerged as a broad area of academic fields and industries. AI algorithms are built with black boxes because of the learning methodology and internal rationality from unanswered questions left by humans about how and why the decisions are made using AI. The sensible and ethical challenge arises due to the absence of explanation results. The evolving sub-domain of AI is “Explainable Artificial Intelligence (XAI)” that affirms implementing a plethora of techniques and tools for unboxing the black box related to AI solutions based on the generation of human-comprehensible, transparent and insightful explanations of AI decisions. The researchers paid less attention to generative models that produce artefacts when compared to producing decisions as outcomes. Concurrently, Generative AI (GenAI) approaches are being applied to the application areas including software engineering. With the usage of question-driven-based and scenario-based XAI design approaches, we aim to explore explainability of the user needs for GenXAI in three software engineering use scenarios: “code translation, natural language to code, and code autocompletion”. Also, this survey evaluates the importance and evolution of GenXAI features for GenAI among diverse applications and domains. Also, this survey includes the classifications and summaries of XAI-based GenXAI techniques along with their applications and practical uses. This research work makes it easier for researchers to understand all explainability over GenXAI methods, and simultaneous accessing of these applications is carried out.