On Philosophy

August 2, 2007

The Functional Properties Of Simulated Worlds

Filed under: Metaphysics — Peter @ 12:00 am

Under normal circumstances there is no reason to object to the idea that simulated objects can have the same properties as real objects, and thus that simulated worlds can embody all the same functional properties as the real world. For example, let us suppose that a real calculator has the functional property of using algorithm A for addition. It has this functional property as the result of its parts having the right causal connections to each other. A simulated version of this calculator would also implement algorithm A for addition. Although its “parts” would really be data in the software doing the simulation they would still have the same pattern of causal relationships to each other, and hence would have all the same functional properties.

Thus the real world and a copy of it running in software have the same functional properties. Let us suppose that the simulation software running such a world works internally by transforming a block of data that represents the contents of the world from one configuration into another over and over again, thus making the simulation “run”. Consider then another piece of software. This one also transforms a similar block of data over and over again, but it does so randomly (truly randomly, not pseudo-randomly). Now, by an extremely unlikely coincidence, let us suppose that the states of these blocks of data, the one that represents the simulated world and the one being transformed randomly, are identical (for the entire time that the simulation is run). The question then arises, does the data brought about randomly have the same functional properties as the data representing the simulated world? Or, more to the point, is there anything it is to be like to exist as part of that data?

The intuitive response is to say that no, it doesn’t have the same functional properties, because the right causal connections aren’t present. In the simulated world the software that is transforming the data follows certain rules, and those rules create the necessary causal relations. In the case of the software altering the data randomly the software does not create those relations, there is no sense in which the state of the data at one time contributed to the state of the data at future times, and so they were completely unrelated.

This answer, regardless of whether it is right or wrong, is not acceptable. It makes causation between different data states dependant on some external force, namely whether the software really is pushing the data around following the appropriate rules or not. But we can’t make that the standard for whether causal relationships hold because then we would have no way to know whether causal relationships hold in this universe. It doesn’t make sense to talk about something “outside” of our universe making events turn out one way instead of another. No, when it comes to our universe we accept that causation is a description of the patterns we find in events, not some additional force that pushes events one way or another. This is the only reasonable position we can take towards causation without running into certain epistemological dilemmas. Thus, by extension, it would seem that whether or not the right causal relationships obtain between various states of the data is dependant only on the existence of patterns in the state of the data over time. And since the data went through the same sequence of states in both cases then we can describe both of them using the same causal description, and hence the same functional properties must hold in both.

Which is not to say that the original assessment was completely off base. What we have here is perhaps two different perspectives on the data. Obviously the software has certain causal dispositions, and the simulation software and the random software differ in their causal dispositions, despite the fact that on rare occasions they may behave identically. This is because the parts that determine their operation participate in more general patterns, and from those we infer that the software does not necessarily behave in the same way. So our intuition that the data associated with the simulation program had the right causal connections and that the one associated with the random program didn’t was really an intuition about the software, one piece of software does have the right causal relationships while the other doesn’t. However, we can also look at the data by itself. The data by itself also has certain regular patterns, which might be described as causal relationships. And these relationships of course depend only on what the data is, not how it was generated.

Now from our external point of view we have a very different attitude about the progression of these two programs. In the case of the simulation we are quite sure that the regularities expressed so far in the data will continue to hold. Our surety is based on our surety that the causal relationships embodied in the program creating those regularities will hold. In contrast, while observing the progress of the random program we will be quite surprised that patterns exist, and we will expect those patterns to be broken eventually. But from “inside” such a simulation things are the same. There is no information about the process manipulating the data, and so an inhabitant of such a simulation will be in exactly the same situation we are, specifically we have no evidence that the causal regularities we have so far observed will in fact continue. Which is one more reason to believe that whether the program generating the data is likely or unlikely to continue creating the same patterns that have so far been observed makes no difference to the functional properties of the data (not the software), since we are in the same position as an observer inside the data, of not being able to know how likely the same patterns are to continue, and since we can’t know obviously it doesn’t make a difference to the current functional properties of the universe (since if it did make a difference then we could know, via that difference, how likely it is the patterns we have observed so far will continue to hold).

Blog at WordPress.com.