ABSTRACT
Epidemiological inference is complex. It is rarely obvious what statistically correlated factors are causally responsible for others. It is typical for multiple possible causes to be viable explanations. Partial evidence is usually misleading. Some of the identified associations normally conflict with expected theory. The complexity of epidemiological inference suggests a coherentist interpretation. Evidence for a claim emerges from a body of information in which the relations of support between claims are bi-directional and may involve rejecting some of the originally accepted claims. To determine overall coherence, the computer program ECHO uses a neural network algorithm for approximately maximizing coherence. ECHO represents each proposition by a unit, a simplified artificial neuron that is connected to other units by excitatory and inhibitory links. As in real neurons, an excitatory link is one that enables one neuron to increase the firing of another, whereas an inhibitory link decreases firing.
