#49 Define and Display Logical Relations Between Items

[Danny Weltman]

This is another underthought feature request, offered in the spirit of perhaps prompting useful thoughts, not because I think this is obviously a good thing to add or because I have any real idea what it would look like if it were added.

It might be nice if, in addition to the main categories of Debate, Position, and Argument (and Works, Persons, etc.) there were a category of Logical Relations. The Logical Relation category would allow me to add logical relations between positions and arguments (and maybe debates). So for instance if argument X entails position Y and ~position Z, I could add a "Logical Relation" describing that entailment. And so on for various logical relations between various arguments and positions.

Then, having defined these logical relations, I would want them displayed in various places. So, just by looking at (e.g.) an argument, I could see its various entailments for various positions.

Right now there is a rudimentary version of this, in that each argument can have various parent positions and for each parent position it can have a verdict. But I don't like that implementation very much, because if my argument is like seven positions deep, and if it has implications for all of the positions it's nested under (e.g. they're all false,) and if I note that in the argument, then it shows up on the tree SEVEN TIMES, which is way too cluttered for me. I've taken to just nesting the argument under the lowest position (or the lowest argument, if it's a counterargument) and then using my brain to tell me that often the argument is objecting not just to the sub-sub-sub-position but to the whole chain. If I could define a Logical Relation to capture that, I could stop just doing that in my head, and I wouldn't have to nest the argument under every single position it is logically related to.

How to display the logical relations is a question. I can think of a relatively easy way to display it when looking at (e.g.) an argument: do it similar to how the argument already displays what it says about the other positions and arguments it's linked to. In addition to listing (e.g.) position X and "False" or whatever, which is the current implementation, it could list "position X" and "false" (this would be redundant, but it would allow me to list all those various parent positions without cluttering the tree) and also more complex things like "position X" and "false if argument Q is true" or whatever.

As for displaying it outside of the argument entries: I haven't thought about this enough to have an answer. Maybe it's hard to show it in the tree. If Hypernomicon ever has any other visualization tools, like an automatically generated mind map or whatever, maybe the logical relations could show up in that. They could anyways should up in the Logical Relations tab, at least, and in the position/argument entries.

Again, none of this is very thought out.

[Danny Weltman]

In case it helps to have a concrete example to keep in mind:

There are a bunch of theories of personal identity. Two of those theories are conventionalism and psychological continuity.

There are a bunch of theories about object existence. Two of those theories are four-dimensionalism and three-dimensionalism.

Some three-dimensionalists are mereological universalists while others are not.

An author argues that the conjunction of (conventionalism OR psychological continuity) and (three-dimensionalism AND ~mereological universalism) is indefensible.

Right now the way this is represented in Hypernomicon is via an argument nested under the "Conventionalism" and "Psychological continuity" views. In the text description of that argument I describe it, including the details that this is only an objection to conventionalism or psychological continuity views insofar as they endorse three-dimensionalism and don't accept mereological universalism.

Cases like this are common, and it would be helpful (maybe? again I haven't thought about this too much) if that kind of information were available in a more formalized and visible fashion rather than just included verbally in the description of the argument.

That is, if the logical relation (C v PC) ^ (3D ^ ~MU) were the thing the argument was listed as rendering false, rather than having the argument listed as rendering false both Conventionalism and Psychological continuity more broadly, that would be clearer.

The more I think about this, the more I realize that a lot of arguments would probably attack not positions themselves, but certain conjunctions and disjunctions of positions. What I've described above is a very common sort of argument: people don't often just attack position X outright, but rather they say that position X won't work unless you accept Y, and many people don't accept Y but some of them do, etc.

[Stephen Mann]

My initial thoughts:

• I agree with the potential usefulness of this feature. One of the first positions I added was Dennett's mild realism. This involves rejecting the dichotomy "beliefs are either real or not real". So I created four position records: (1) Beliefs are real; (2) Beliefs are not real; (3) Beliefs are either real or not real; (4) Mild realism. Clearly these positions are logically related. I can imagine eliminating (3) and adding logical relations between Mild realism and the other two positions.

• There is a question to be asked about best-practice: should we create position records for both P and not-P, or just one? (I have a memory that this was raised here but can't find it now.) A feature like this would impact that issue.

• I agree that the tree can get very busy when all positions/arguments are explicitly linked to all other positions/arguments they are logically related to. To the extent that the proposed feature would ameliorate that, it would be a good thing.

• I agree that some kind of mind-map/directed graph is the most intuitive way to display these relations. The only relevant example that comes to mind is the causal modelling visualisations of CHIELD: https://correlation-machine.com/CHIELD/document.html?key=sterelny2012language

I will think more on this and add comments if/when I have them.

[Danny Weltman]

The discussion about position records for P or for P and ~P is here: https://sourceforge.net/p/hypernomicon/discussion/general/thread/b21057ca13/#c3b6

This feature would solve the issue for me, I think, since I'd just put position P in, and then ~P would be a logical relation. So, whatever I would've previously put in the position record for ~P, would instead go in the description of the record for the logical relation ~P. That would be what I would do for cases where ~P is not a substantive enough thing to warrant its own position. For cases where ~P is substantive enough to warrant a position, maybe the logical relations could be linked to positions the same way terms can be linked to positions.