Belief revision is constrained by cognitive limitations, with evidence alone often failing to drive change. Research suggests that resistance to epistemic change is not necessarily irrational but stems from the need to maintain internal coherence within complex belief networks. When confronted with conflicting information, individuals employ strategies like modifying auxiliary hypotheses to preserve coherence rather than overhauling their belief system entirely, consistent with the Duhem-Quine thesis. This approach minimises cognitive effort, as belief revision is resource-intensive, and ignoring evidence also carries cognitive costs. I study how interventions shape belief network formation and revision, focusing on trade-offs between accuracy, coherence, and cognitive effort. This project examines how belief networks form and evolve, investigating the trade-offs between maintaining internal consistency, aligning with external evidence, and managing cognitive demands. Through experimental studies, we explores the conditions under which people prioritise coherence over accuracy, how cognitive load influences the restructuring of belief networks, and whether there are thresholds that determine when minor updates, core adjustments, or complete revisions occur. A computational approach helps formalise these processes, modelling belief revision as a sequential local search within a single causal framework. These studies aim to advance understanding of how individuals integrate new information while maintaining functional belief networks, and have broad implications for understanding the persistence of misinformation, resistance to epistemic change, and the mechanisms underlying rational belief maintenance in the face of conflicting evidence.