Bounded Rationality and AI in Decision Making

Abstract

Purpose: This paper explores the intricate relationship between human decision-making and artificial intelligence, viewed through the lens of Herbert Simon’s bounded rationality theory. It investigates how AI simultaneously augments and constrains human cognition, while raising urgent ethical, environmental, and governance concerns that remain inadequately addressed in existing literature i.e. whether AI genuinely frees human cognition from its inherent limitations, or merely substitutes one set of constraints with another while generating fresh ethical and governance challenges in the process.

Design/Methodology/Approach: The study builds on a systematic review of over 300 sources spanning behavioural economics, cognitive psychology, AI ethics, neuroscience, and organizational theory including peer-reviewed journals, government reports, and industry white papers. The review follows the arc from Simon’s original challenge to the myth of homo economicus, through Kahneman and Tversky’s work on heuristics, and onward to AI’s journey from early rule-based systems to today’s deep learning models operating across healthcare, finance, and manufacturing.

Findings: AI does not overcome bounded rationality, it relocates it, shifting constraints from the human mind to computational systems. In complex domains, AI sharpens decision quality considerably. Yet it also breeds new forms of irrationality: entrenched algorithmic bias, quiet erosion of critical thinking, corporate-engineered choice environments, and platform designs built more around engagement than genuine user benefit. Where corporations hold the reins, these tendencies tend to deepen existing inequalities rather than correct them.

Originality: The paper puts forward the idea of AI-influenced rationality — arguing that AI does not sit passively beside human judgment but actively moulds it. This perspective bridges a genuine gap in the literature, connecting Simon’s foundational theory with the lived realities of modern AI in a way that existing frameworks have not.

Paper Type: Review of Literature.