Yes. When the mechanism keeping the trigger from inadvertently springing the trap after it is set is merely the static friction of a tangential force to the direction of the trap force, we have the below, where $F$ is force, $W$ is work, $x$ is the direction of the trap movement, $y$ is the direction of the trigger movement, and $k$ is the mechanical advantage gained.
W_T = \int F_x \, dx \\
W_t = \int F_y \, dy \\
W_T = k \, W_t
It is not unusual for the friction of the trigger device to be deliberately reduced so that $k \ge 100$, meaning that the energy transferred from the animal to the trigger is no greater than 1% of the energy unleashed on it.
That the system has
- A detection device (the contact point of the trigger),
- A control function (the identity function),
- Energy storage (the consequence of the setting action), and
- A motor function (the trap release)
These qualify early traps not only as machines but as automatic one. The human that set the trap does not perform the identity function. The triggering design was performed in advance by humans, and even the purchaser of the trap need not understand it to set it and capture an animal. This clearly fits the formal definition of automation.
No. The identity function does not require computation.
Deliberate Information Obscurity
This is not the only historical or cultural aspect to the information science involved. That the existence of the trigger and the action it triggers must be unknown to the animal is part of the information science involved. The secrecy is part of the mechanism, and the secrecy may not involve hiding the trigger or the spring mechanism from the animal. That the animal does not understand mechanical engineering may be ample secrecy for an animal that has not learned in DNA (by repeatedly eliminating lines of offspring that could not visually discern that something was amok) or
Does the System Learn?
No. It does not improve each time it is used. However, learning traps have been designed and are deployed today. The pray is the consumer in the case of marketing automation. The pray is the cyber-enemy in the case of national infrastructure defense or attacks upon it.
Is Any Learning Involved in Mechanical Traps?
Certainly. For the earliest traps to be fabricated, much had to be learned in advance by humans, again, either parametrically in DNA or in the human brain through experience or through language and education. These are just a few of many cognitive transitions that had to occur, represented as sentences.
- Animals can be killed or maimed and then used as food.
- Animals can sense human presence and flee.
- Bending and releasing wood can propel somewhat like an arm can.
- The releasing can be suspended
- The animal can cause the release unintentionally
It is interesting to note that to artificially recreate the above learned understandings, both DNA and neurology were involved but not gradient descent, because the experiments and the analysis of results were discrete and no gradient could be calculated. It is also unlikely that it was entirely trial and error in searching for better ways to get fed. This is why logic based AI continues to be an important area of research.