OK, so I want to jump into the socionics test bandwagon. I want to combine the best aspects of the ones out there, have it be interesting( interesting as it could be), and have it be accurate. I would also like help. Or just to talk out the ideas in my head for it.
My initial points are:
1. Ask the questions most likely to get the most accurate answer.
2. Don't over think it.
3. Use tons of possible testing techniques.
4. Keep it random so it's a new order of modules every time you take it, as well as asks (some of the time) random questions from each module.
5. Track results from a single entry point, so you can get take a rest and come back to the exact same place at any time.
6. Each test question can be assigned a value(metric) according to it's likelyness to be accurate. It'll have to be a metric or metric group going only to a final type(ENTP, ISFJ, etc).
My initial work will be with designing a scalable infrastructure to handle these requirements. Next will be starting it off simple. Then, I can create these 'modules' of tests and add them into the program. Initially I will have these four dictotomies tested:
each of those numbers will be a module. For instance, you will have a screen with two choices (on random sides of the screen), pulled from a random sampling of test questions (which is generated random but becomes static), in order of modules (generated random, becomes static).
The choice you make(let's say you clicked the options associated with being introverted) will add -1 to the 8 extroverted types, and +1 to all of the introverted types. Same for each module. At the end, the result program is run and you are presented with the test's results in the manner of a type, and the runner up type.
Then I plan to add the following modules:
1. Socionics mood descriptions (as shown on the16types.info)
2. Likes and dislikes from the Oldham associations of the types from ptypes website
3. descriptions similar to Hugo's tests, where like a positive on an ESFJ description will be super negative score to the conflict(INTP).
4. Skills - Basically the functions ( ) part of the test. This will be last because it will be complicated to assign metrics going towards a type.
What I have so far will be available soon.