I picked up Dependence Logic: A New Approach to Independence Friendly Logic by Jouko Väänänen. I figure I’ll write up a review when I am finished with the book, but there is one chief difference between Dependence Logic and Independence Friendly Logic that needs to be mentioned.
On pages 44-47 when describing the difference between Dependence Logic and Independence Friendly Logic Väänänen says,
The backslashed quantifier,
∃xn{xi0,…,xim-1}φ,
introduced in ref. [20], with the intuitive meaning:
“there exists xn, depending only on xi0,…,xim-1, such that φ,”
…
The slashed quantifier,
∃xn/{xi0,…,xim-1}φ,
used in ref. [21] has the following intuitive meaning:
“there exists xn, independently of xi0,…,xim-1, such that φ,”
which we take to mean
“there exists xn, depending only on variables other than xi0,…,xim-1, such that φ,”
The backslashed quantifier notation is part of what Väänänen calls ‘Dependence Friendly Logic’, and is equivalent to the ‘Dependence Logic’ that the rest of the book expounds. This backslash notation makes the difference between Dependence (Friendly) Logic and Independence Friendly Logic clear by showing that the former logic takes the notion of dependence to be fundamental whereas the latter takes independence to be fundamental. Väänänen takes this to be an advantage because he says that Dependence Logic avoids making
one ha[ve] to decide whether “other variable” refers to other variables actually appearing in a formula ?, or to other variables in the domain…
However, this treatment misses an important philosophical difference between Independence Friendly Logic and Dependence Logic. Dependence Logic is fundamentally based upon Wilfrid Hodges work, ‘Compositional Semantics for a language of imperfect information’ in Logic Journal of the IGPL (5:4 1997) 539-563, in which Hodges lays out a compositional semantics for languages such as Independence Friendly Logic using sets of assignments instead of individual assignments to determine satisfaction (T or F). Väänänen infers that Independence Friendly logic is just a bit unruly when it comes to specifying variables because he is working within a system that assumes sets of assignments are a useful and unproblematic way to determine satisfaction.
However the unseen problem of using sets of assignments is that something is added by assuming the domain is a set. For example, let’s take try to define a location and take the set of all the points in the universe. However, we immediately run into relativity: All locations are defined relative to each other and the people trying to figure out where things are, i.e. There is no predetermined set of all the points in the universe. The issue is that the domain of potential assignments, the objects in the universe, may be dependent upon the person or people using them (the players of the semantic game in this case). If the domain is dependent upon the players, the set cannot be constructed until after the players have begun the game. Therefore, if we postulate that the domain is a set at the outset then the players know something about the game that they are playing, namely that it does not depend upon them because it was predetermined.
Following this line of thought it seems possible to constructed a game in which the domain {Abelard, Eloise} is such that Abelard and Eloise are the actual people playing the game and the formula is ‘Someone x lost the game by instantiating this formula’ such that whoever instantiated that formula would win the game according to the rules. But then the formula would not be satisfied, so that player would have lost, but then it would be satisfied, a paradox. It is easy enough to declare that the domain must be independent of the players, but again this signals something about the game being played to the players before the formula to be is revealed.
Lastly there is something to be said about using logic to represent natural language here too: if you consider the set of all possible responses to some question, you are not ever considering all possible responses, but all the possible responses you can think of at that time. Therefore if we are using game semantics and imperfect information to represent natural language, then it is a mistake to predetermine the domain of all possible responses separate from the people involved. Again, the domain being linked to the people involved is at odds with the domain being a predetermined set.
Long story short, there is a very good reason for not always using sets of assignments to determine satisfaction. Depending on the situation, a set may offer non-trivial information about a game or misconstrue the game being played. Independence Friendly logic makes no assumptions about the type of game being played and is therefore of greater scope than logics that are based upon Hodges work. Of course one is free to use sets of assignments to determine satisfaction and derive set-theoretic results, but the compositionality gained comes at the price of limiting the types of games that can be played.
When I decided to write this post it was a bit over a month ago. The only problem was I didn’t have Hodges paper so, though I could infer what he said, I didn’t actually have the reference. And since I’m just some random fool running around New York with no academic affiliation, getting a-hold of an academic journal turned out to be quite the problem. The only ones with direct access to the Journal of the IGPL V.5 #4 was NYU and they are stingy about letting random scary people into their libraries. However, they do allow you to text message their librarians and they sent a message back telling me to go to the New York Public Library. I had looked on the NYPL site and it did not list the correct reference, but I figured that I possibly had screwed up, so I went.
The NYPL did not have the correct reference: they did have some of the IGLP Journal, but not what I needed. But they said they could acquire it for me. First they called NYU, who they said was stingy and didn’t like to help out , and got the automated phone message run around. So then they said they would order it for me from some academic institution. It would be copied/printed out and mailed to me at no charge, most likely from UPenn, which was the closest obliging institution.
That copy arrived yesterday (5/27/08). So cheers to the NYPL. A big thumbs down to NYU for making it difficult for non-academics to do work when they are the only ones with the references, and much thanks to my friend JC who goes to NYU and was able to get me the paper in about 5 minutes when I was smart enough to ask him to access it for me.