Integrating dependent evidence: naïve reasoning in the face of complexity

AbstractWhen reasoning about evidence under conditions of uncertainty, one important consideration for accurate updating is the presence (and influence) of dependencies. For instance, if considering whether a patient has a disease, the value of two doctors’ diagnoses indicating the presence of the disease may carry more value if such diagnoses were conducted independently, rather than if, all else being equal, one doctor has seen the other’s diagnosis before making their own. In the present paper, we demonstrate that lay reasoners prefer to avoid dependencies when considering evidential support. However, we additionally illustrate two cases in which dependencies may carry evidential advantage: namely, when information is partial or contradictory. Lay reasoners erroneously remain averse to dependencies even in such cases, reflecting the difficulties inherent to considerations of dependence.

Return to previous page