called the utility of the prospect Oj. The properties of this function are: 1. u{01) u{0z) if and only if the individual prefers 01 to 02. 2. If 0 is the prospect where, with probability p the individual faces Oi and with probability i-ft he faces 0%, then u(0) ft u(0i) x-ftu(0f). Much has been written about the concept of utility, see e.g. the historical and critical comments reviewed by L. J. Savage in [9]. A very clear exposition is also given by R. Duncan Luce and H. Raiffa in [5]. It seems very difficult indeed to assign numbers to prospects, measuring their desirability and taking account of financial losses or gains, personal satisfaction, ethical values etc. connected with each prospect. The originators say (of models in general)"Similarity to reality is needed to make the operations significant". We shall later return to the difficulties connected with this requirement in the context of geodesy. 3. The Elements of a Decision ftrohlem We shall introduce the different elements of a decision making problem by means of a simple example. Suppose a surveyor is measuring some distance by tape. He makes a direct and a reverse measurementthere will be a difference between the two measure ments, which we will denote by t. From experience he knows that under the prevailing circumstances the measuring procedure results in a standard deviation of 2 f2 cm in a single measurement, and that the two measurements are stochastically independent. Now there may be a gross error in one of the measurements. For reasons of exposition we make the very much simplifying assump tion that he has either made a gross error of 10 cm in one of the measurements, or no gross error. These are the two possible states of nature. The only indication of the possible presence of a gross error is the size of t, which is a realization of the observation variate t. The determination of t (or in fact the second measurement) can be seen as an experiment to get information on the state of nature. There are two actions he can take, after having found t: remeasure or not remeasure the distance. Each of these actions will have certain consequences: remeasuring takes time, but enables him to remedy a gross error if he has made one. If he has not made a gross error, his time will be lost. On the other hand, if he does not check the distance once again, his result may be false. These consequences are reflected in the losses of utility associated with each action under each state of nature. Our surveyor will let it depend on the size of t whether he will remeasure or not. For certain values of t he will decide to remeasure, 124

Digitale Tijdschriftenarchief Stichting De Hollandse Cirkel en Geo Informatie Nederland

Tijdschrift voor Kadaster en Landmeetkunde (KenL) | 1967 | | pagina 6