Reliability and Imprecision: FEW 2005 comments on Haenni and Hartmann

When we say of something that it is unreliable we are not necessarily saying that the chances of it behaving as one expects equal the chances of it behaving otherwise. The train service in Lisbon is reliable, I tell you, yet the ambulance service in Lisbon is unreliable. By this, I am not saying that I’ve never boarded a late commuter train in Lisbon, nor am I saying that it is a toss of a coin whether dialing 112 will fetch you an ambulance. Indeed, you would find it surprising were this to be what I meant by ‘reliable’ and ‘unreliable’. Moreover, if this was what I believed to be to true of Lisbon trains and ambulances, you’d think I was understating each case by putting it this way. Either way, you would find both the random sense of unreliable and the “with certainty” sense of reliable extreme. Yet, this is the concept of unreliability used in the Haenni-Hartmann (HH) model. In their introduction of reliability variables, REPi, indexed to an information source i, Haenni and Hartmann write that