The Myth of Hybrid Models in Artificial Intelligence: Are They Truly Better Than Individual Models?

In more detail, hybrid models of artificial intelligence are currently thought to offer better performance compared to standalone models by leveraging more than one paradigm of machine learning within a single framework. In recent years, this opinion has shifted from its initial hypothetical stage to becoming more of an assumption within the research as well as applied fields of artificial intelligence. This paper critically evaluates the validity of this assumption by analyzing available research studies, limitations of theories, as well as observations based on available evidence from various fields, such as medical diagnostics, finance, as well as public decision-making. Conclusions derived from this paper confirm that hybrid models are unable to offer better performance compared to well-designed standalone models, which largely depend on context, data, as well as models of assessment. In additional discussions, this paper also claims that overemphasis on unrealistic assumptions regarding models might hinder transparency as well as offer useless confidence.