6.2 Constraint
63
accounts of major events such as military victories were preserved in massive stone
monuments whose inscriptions can still be read today several thousand years later.
Military secrets are printed on paper or photographed using silver halide film and
stored in bunkers, rather than committed to magnetic media. We tend to write down
things we need to remember for a long time.
The value of information is closely related to the problem of weighing the credi-
bility that one should accord a certain received piece of information. The question of
weighting scientific data from a series of measurements was an important driver for
the development of probability theory. Daniel Bernoulli (1777) raised this issue in
the context of averaging astronomical data, where it was customary to simply reject
data deviating too far from the mean and weight all others equally. 22
Bennett has proposed that his notion of logical depth (Sect. 11.5) provides a formal
measure of value, very much in the spirit of Eqs. (6.26) and (6.27). A sequence of
coin tosses formally contains much information that has little value; a table giving
the positions of the planets every day for several centuries hence contains no more
information than the equations of motion and initial conditions from which it was
deduced, but saves anyone consulting it the effort of calculating the positions. This
suggests that the value of a message resides not in its information per se (i.e., its
absolutely unpredictable parts) nor in any obvious redundancy (e.g., repetition), but
rather in what Bennett has suggested be called buried redundancy: parts predictable
only with considerable effort on the part of the recipient of the message. This effort
corresponds to logical depth.
The value of information is also related to the amount already possessed. The
same Bernoulli asserted that the value (utility in economic parlance) of an amount
mm of money received is proportional to log left bracket left parenthesis m plus c right parenthesis divided by c right bracketlog[(m + c)/c], where cc is the amount of
money already possessed, 23 and a similar relationship may apply to information.
6.2.2
The Quality of Information
Quality is an attribute that brings us back to the problem posed by Bernoulli in 1777,
namely how to weight observations. If we return to our simple measurement of the
length of a piece of wood, the reliability may be affected by the physical condition
of the measuring stick, its markings, its origin (e.g., from a kindergarten or from
Sèvres), the eyesight of the measurer, and so forth.
22 See also Euler (1777).
23 Bernoulli (1738), cf. Thomas (2010).