Artificial computation is increasingly critical to the function of our society. Popular conceptions of computation may be somewhat circular, expressed as whatever computers or digital electronic circuits "do.\ From this perspective, there are some very useful but philosophically trivial things that computers "do," such as refreshing the state of data in storage media, or copying data. Treating computing arbitrarily as performing operations upon data also yields little insight. Importantly, computation is what our nervous systems are doing (with significant analog processing and noise), and I would argue that the operation of any physical system may be understood as computation as well. A surprisingly common mistake among scientists, who should know better, is to equate computation with computability.
In the book Reliable Computation in the Presence of Noise (1963), authors Shmuel Winograd and Jack Cowan offer a surprisingly little-known, but strategically useful, definition for computation: the selective reduction of information across a channel (a communication channel or a processing channel).
What does this mean?
If I add two numbers together, the sum preserves information only about the total quantity, which was always there in implicit form, but information about the original addends is lost in the sum.
If I make a yes/no decision based on looking at an image (e.g., does it contain a human face?), the decision reduces all of the information in the image to a single bit.
I may apply a wide variety of forces to the steering wheel of my car, but it is engineered to only respond by rotating around its axis. The steering column only transmits force information that corresponds to its rotational sensitivity; the rest is lost.
Regulatory functions like homeostasis principally operate by damping or correcting perturbations from optimum conditions.
From this point of view, engineering functions into physical systems involves selectively tuning their information reduction properties.
The author's affiliation with The MITRE Corporation is provided for identification purposes only, and is not intended to convey or imply MITRE's concurrence with, or support for, the positions, opinions, or viewpoints expressed by the author.