The precision of DNA identification with a certain locus set is typically measured with the frequency P of allele combination
occurrence in a population. If the decision is made solely on the basis of DNA analysis results, it would always be the nonzero
probability of justice failure due to mistaken identity. Therefore, for every case in hand we have to provide rationale for
the critical value Pmax of identification precision which can be assessed as sufficient for the judgment. We employ the notion of
the average utility from the decision theory to obtain the closed form expression Pmax=ln (1+x)/(V-1) where V is the estimate
of the number of potential suspected persons and x=(a+b)/(c+d). Here ‘a’ is the utility gain when an offender is convicted, ‘b’ is
the utility loss when an offender is acquitted, ‘с’ is the utility loss when an innocent is convicted by mistake; and ‘d’ is the utility
gain when an innocent is acquitted. The value of ‘x’ reflects the adoption level by the society of justice failure in the considered
case or a group of similar cases. The adoption level is a function of the society which needs to be elicited from a special poll.
A possible question of the poll is the following decision problem: “Imagine you definitely know an innocent was convicted
as a result of a justice mistake. You can secure an acquittal to this accused person but several convicted criminals (certainly
guilty) will also be released from prison. How many criminals (1, 2, 5, 10 etc) would you release to save the innocent”? Then
the maximum number ‘n’ of criminals released gives x=1/n. We provide examples of calculation of the critical frequency Pmax.