Robotics Reference
In-Depth Information
Robot Feelings
Could a robot have feelings? Some say of course. Some say of
course not. [11]
On a subject such as this one, philosophers can have a field day. In his
1959 article, “The Feelings of Robots”, the Harvard professor Paul Ziff
raised a number of objections to the concept of robots having feelings
and he came down squarely against the proposition that they will. But
most of his arguments fail to hold water, not only in my view but also in
those of some of his philosopher peers. One of Ziff 's arguments is that
Robots may calculate but they will not literally reason. Perhaps
they will take things but they will not literally borrow them. They
may kill but not literally murder. They may voice apologies but
they will not literally make any. These are actions that only persons
can perform;
ex hypothesi
robots are not persons. [11]
Hole number one: why are these actions that “only persons can per-
form”? Ziff 's rationale presumably relies on a definition of “action” that
incorporates the intention to carry out the action, and if robots do not
have intentions then, presumably, Ziff would argue that they can not
carry out actions. But if a robot takes your hat off your head, leaves
the room for a few minutes and then brings it back, replacing it neatly
on your head, then the robot has had your hat temporarily. Try asking
Google for a definition of “borrow”, and top of the list is “get
temporarily”—the robot has indeed borrowed your hat. And since it
carried out the action of borrowing your hat the robot must, if we follow
Ziff 's thinking, have intended to do so.
Another of Ziff 's arguments is his criticism of fellow philosopher
Donald MacKay's observation, that any test for mental or other attributes
that needs to be satisfied by the observable activity of a human being can
be also passed by a robot. MacKay asks, “What would be wrong with
the robot's performance?”, to which Ziff responds by saying that nothing
need be wrong with it; what is wrong is that it
is
a performance. In anal-
ogy with the Turing Test I contend that, if a performance is so convincing
that one does not realise it
is
a performance, then one should accept the
performance as being humanlike, and that the performer has passed the
test and possesses whatever attribute the test is designed to detect. So
in this respect Ziff 's performance argument flies directly in the face of
Turing.
