next up previous
Next: Axiomatization Up: Cartesian Counterfactuals via State Previous: Cartesian Counterfactuals via State

Relating State vectors to Propositions

The above way of rephrasing a counterfactual as a statement about state-vectors is insufficient in general. It does not specify how each step is to be done in other cases. There are three steps.

  1. The first is to describe the current world as the contents of a state vector. This is equation 10 above.
  2. We now relativize all the other terms, those that are not co-ordinates, in the theory, making them functions of a state vector tex2html_wrap_inline1464 . We replace the terms t that are co-ordinates by their the contents functions applied to the term, i.e. c(t,xi). This gives the equation 11 above.
  3. The third and final step is to interpret left hand side as a modification of the state vector. This step yields equation 13 in the above example. We then judge the relativized right hand side applied to the new state vector. That is we ask if the relativized right hand side follows from the values of the co-ordinates initially, and the relativized theory.

Our state vector does not have values for all terms, just those in our co-ordinate frame. Thus, for our numerical example, the set of co-ordinates is,

We need to distinguish between the value of a constant x and its name tex2html_wrap_inline1516 . Here tex2html_wrap_inline1516 in bold font refers to the name x, while x in plain font denotes the value of x. Let the state vector tex2html_wrap_inline1464 assign these the termsgif tex2html_wrap_inline1528 . The propositions that the state vector tex2html_wrap_inline1464 encodes are thus the three propositions,

Here we have used x, in plain font, as the proposition states that the value of tex2html_wrap_inline1516 is the value of the term tex2html_wrap_inline1536 .

We assign a co-ordinate a value using the a assignment function. We can determine the value of a co-ordinate in a state vector using c, the contents function.

We call the set of co-ordinates our frame, and we assume that they are exactly the names of terms that are true of the predicate f. Rather than assume that there is only a single frame, we can make the frame an argument of c and a. This gives the axioms,

However, for notational convenience we assume that the frame is recoverable from the state vector, so we keep the ternary a and the binary c whenever what frame we are using is clear from the context.

Our state vector thus uniquely determines the values of the co-ordinates in the frame. From these values, we can determine a theory, the set of sentences implied by the statements that the co-ordinates have the values in the state vector.

The propositions that our state vector gives us may not be all the facts about the world. In our numerical example, the extra fact,

was also the case.

We usually insist that our approximate theory is consistent with our frame. That is, the propositions true at every point in our frame are consistent with the approximate theory.

To reiterate, a co-ordinate frame, f, is a tuple of names of terms. In the simplest case these are constants. A point p in a co-ordinate frame f is a tuple of terms, equal in number to the co-ordinates of f, and having the type/sort of the corresponding term in f. In the simplest case, this is a tuple of values, the denotation of the terms. The state vector that assigns the co-ordinates f the terms p is denoted tex2html_wrap_inline1566 .

The relativization of a theory A to a state vector variable tex2html_wrap_inline1464 , written tex2html_wrap_inline1572 and a set of co-ordinates X is the theory resulting from replacing each co-ordinate x by , and replacing each function, predicate and constant k not in X by a function from tex2html_wrap_inline1464 to the type/sort of k.

definition298

We now consider our first example. We define our frame as follows.

It is important to note that tex2html_wrap_inline1604 are names of constants, not variables themselves.

We now define our initial state vector tex2html_wrap_inline1480 .

We need unique names axioms for the names of terms. We use natural numbers as their own names, for simplicity.

We then have that

We can derive this from the second property of a and c, and the fact that tex2html_wrap_inline1516 tex2html_wrap_inline1614 tex2html_wrap_inline1616

We have x = 1, y = 3, z = 1 tex2html_wrap_inline1620 tex2html_wrap_inline1622 .

We have this set of sentences as our theory, as these are the formulas that result from the terms x,y etc. to their values, 1, etc. in the state vector.

We now prove that the first two counterfactuals that we considered in our skiing story are true and false respectively. To show this we need to introduce an axiomatization of skiing, which we do in the nest section. The axiomatization is only needed for proving the later theorems, and may be skipped by the un-interested reader.


next up previous
Next: Axiomatization Up: Cartesian Counterfactuals via State Previous: Cartesian Counterfactuals via State

John McCarthy
Wed Jul 12 14:10:43 PDT 2000