Philosophy 105

I’ll go through the two models to refresh your memory before I go on:

USER MODEL:

1) M performs T
2) A uses M to perform T
3) A is responsible for T

Because the technology is “transparent” to the agent – the machine is only an instrument, while the agent is autonomous – the agent is responsible for the task performed by the machine.  This is how most technology is explained; technology cannot act directly, and can only perform when incited to do so by an agent.  Because most people realize that most (if not all) of the machines we use with any regularity are not autonomous, the “agency transparence” or immediate elimination of the machine as a candidate for responsibility is a natural result.  In the case of machine malfunction, we cannot blame the agent for any negative result and must, instead, point to an error in design.  Hence:

DESIGN MODEL:

1) M performs T
2) A designed M to perform T
3) A is responsible for T

Here, the user’s intention is irrelevant in light of the designer’s intent.  If one hard-wires sticks of dynamite to a car’s engine in order to detonate it once the key is turned, you cannot blame the man who turns the key for blowing himself up.  Clearly, the man who engineered the explosion is at fault, and that is how we would expect the case to be handled in court.

The difference between the two is simply one of intent.  If the user did not expect M to perform a specific T, the designer who enabled M to perform that unexpected T is the one to whom you would attribute responsibility for the result.  If the user willingly allowed M to perform T, the designer is irrelevant; you cannot blame the engineer behind the gun’s function for any murders involved except those that the user did not intend to perform.

Dan says that “artificial intelligence calls into question the idea that technology is an extension of human agency,” and I am forced to agree; what AI does is allow allow M to have agency, or act as an A.  Rather than a human pulling a trigger, imagine a Star Wars battle droid doing so; in that case, the droid would be responsible for the gun’s firing and the death that might follow.  But wait!  The droid is, despite being artificially intelligent, still a machine.  It, therefore, had a designer who designed the droids to perform their deadly tasks.  Who, then is responsible?

This gray area extends to deep blue, recalling the design model from class:

1) Deep Blue bested Kasparov at Chess
2) IBM designed Deep Blue
3) IBM bested Kasparov

Even granting that Deep Blue is an intelligent machine, its ability to defeat Kasparov was completely bestowed upon it by IBM.  This is different than someone teaching a child to perform a task because there is some effort or ability required on the part of the child.  Deep Blue’s formidable chess skills are no more a result of its inherent skill than a really excellent nose is the result of one’s proteins if they’ve recently undergone rhinoplasty.  It seems as if there is something artificial that must be present in the machine that cannot allow the machine to take full responsibility for its actions.

Largely, I think this is because of game theory.  There must be a better method of demonstrating artificial intelligence than defeating even the best humans at games.  Almost every video game has a “player vs. computer” mode programmed into it.  When I was younger, my mother would occasionally humor me and play some NES game with me.  Inevitably, when confronted by even the easiest of computer settings, she would lose.  Even if she were very famous, I doubt anyone would attempt to classify my NES as artificially intelligent.  That’s a simple example, but the idea is similar: machines are very, very good at performing algorithmic tasks.  They are much better at this than humans, and inevitably will best humans at things involving that sort of reasoning.  Games can become increasingly complex, but if a computer is able to select the best move from an increasingly large number of possibilities at a faster rate than we can, we expect it to win at least some of the time.  The programmers give the machine a set of values, refine them through testing, and stomp the competition.

I’m going to stop writing here.