Probably the first constructive sit down run through of the concept that I have had since ‘joining’ the project. I mean joining in a loose sense seeing as it has mainly been an idea slowly growing inside Douglas Kelly’s head for who knows how many years. We’ve spoken of it fairly often over the past 12 months and with ‘Synergy’ basically completed the time has come to finally start concentrating on bringing Wikilogic to life.
There was a planning session three weeks ago to work out where and how to begin. We have split the tasks into two main components: Tech and Philosophy. The Philosophy has to come first, if this concept doesn’t work out there’s no point in looking at the tech. So the first task: research. Reading reading reading! This applies more to myself as I have a much smaller grounding in the subject of argument theory, sure I’ve read a few books and chatted in cafes about this stuff but I’m way off the required knowledge for helping to found a project like Wikilogic. Douglas has had an interest in this subject, not to mention the original idea, for years now. He is the resident expert, for now!
As of writing this it looks like the most promising book we have found is ‘The uses of argument’ by Stephen Toulmin. This will have to be the subject for a later post, I’ve been firing ahead with learning the various technologies that go into Web Development. So skipping over that initial research phase, we are brought to tonight.
The three of us (Douglas, his girlfriend Svenja and myself) sat down with some paper and decided to work though a statement: ‘crisps are healthy’. To visualise the progression of the supporting arguments I roughed out a ‘shower time’ idea that had been rattling around upstairs for a few days. Write the statement, surround it with a box and draw two smaller internal boxes, one at the bottom left and the other at the bottom right. These represent the supporting or contradicting statements. So, to work with the chosen example. Another statement is added in it’s own box below our original, ‘crisps are potatoes’. This is linked by a line to the bottom left corner box of the ‘crisps are healthy’ box. Anything linking to the bottom left indicates that the connected statement must be true in order for the original statement to be true. Below ‘crisps are potatoes’ we added another in the same fashion: ‘potatoes are healthy’. This is as far as my small idea went, and already there is a problem.
The simplicity killer. Well, not quite as it turns out. The problem was that the hierarchy of the two connected statement could be swapped. What was happening was that both were required in order to have any relevance to the original statement. This means they are of equal value. Here we edited the diagram layout. instead of a single connection from the original statement to one other statement, we group the two supporting statements and link the original to the group. So, if any statement in the group is disproved, the connection from the whole group breaks and the original ‘crisps are healthy’ becomes false.
Now to take this a little further. To the bottom right internal box of the original statement we add another group of 2. ‘crisps are fried’ and ‘fried food is unhealthy’ (don’t worry about the actual truth of these statements, this is something wikilogic will fix in the future when they are extended to their own supporting statements.) So, anything connected to the bottom right internal box means that the connected group must be true in order to disprove the original statement. This will lead us onto our biggest problem but first, another problem: mirroring arguments.
Imagine in one corner of the internet the first statement is made, ‘crisps are healthy’ and the first supporting group is connected ‘crisps are potatoes’ and ‘potatoes are healthy’. Then in anther corner of the internet another statement is made ‘crisps are not healthy’. This leads us to apply a rule: only statements of truth may be added, not statements of something not being true. Now, the statement can be reworded: ‘crisps are unhealthy’. This can be linked in the truth box (bottom left) to the second group of two: ‘crisps are fried’ and ‘fried food is unhealthy’. Now, someone notices that these two original statements are actually mirrors of each other in which case they must be linked.
This mirroring of statements is not necessarily a breaking point, you can literally flip one box over and attach it to the back of the other and the two kinds of dependancies will match up. The problem starts to arise when you think of a group of statements like: ‘Yahweh is the one true god’, ‘Allah is the one true god’ and ‘(Christian)God is the one true god’. These are also mutually exclusive but we now have a 3 way mirror and, as you can probably imagine, this may extend to any number of claims so we can effectively have an infinite mirror. Yikes, how are we going to work this out in an easy to use interface for the general public? Well, turns out that this slightly wooly problem wasn’t actually the root of our issue. We talked it all over and it came down to what we must so with what you might call ‘root statements’.
One thought is having a default state. When a statement is made claiming a truth, it is automatically set to false until proven true. The problem here is that when you connect a supporting statement in order to prove it’s truth your supporting statement is automatically set to false. You must now prove you supporting statement, and so on. We discussed the idea of allotting the probability of an end node statement (or leaf statement) and having the values of all the ending statements propagate up the tree, but, this is starting to introduce a lot of complexity into the system and the idea of something being hazy / grey doesn’t help our aim out much. But, it turns out this issue with what to do with root statements seems to have it’s own Wikipedia page, how lucky!
After discussing and defining our problem we broke off in need of rest, to get to this point had taken several hours! I began randomly googling fun philosophy stuff, a list of unsolved problems. Starting with moral luck we had a read and a wee discussion then, out of sheer luck, I clicked on Infinite Regression as it looked interesting. Have a read of the article and you might well imagine our surprise and delight at having stumbled upon the proper definition of our problem. And, through it, a path that traced as follows: Coherentism > Difficulties for coherentism > Bayesian probability.
I think our minds were collectively blown at having potentially found a solution to our problem. A little digging around the subject area brought up Frequentist inference, Fiducial inference and Statistical inference (the overarching concept). It is at this point that we left off. Looking into how this works and whither it is actually a possible solution will be for the next meeting.
What a start! Wikilogic is starting to roll.