Developing an accurate test - put your suggestions

[machintruc]

I'm currently thinking of an easyTIM-like project.

I want this test to :

- have easy to reply questions, that are interpretated quite the same way by most people
- have questions that produce very limited self-image-like lying
- output the correct TIM at least 70% of the time, even if it needs very complex algorithms
- evaluate the physiological profile, to determine the variant it's useful especially for LIE's.

I'm currently thinking of an "hybrid test" i.e. there'd be an hardware part and a software part.

1. You fill a physical paper form
2. You input your answers
3. The answer is output - you may print it

But why paper ?

- Computers crash, not forms. Even if it's under pure MS-DOS or such, a power cut may happen.
- It's easier to do the test partially and to resume it later.
- You don't have to monopolize a computer for somehow 40 to 90 minutes, but only for 3-4 minutes.
- If you lose the results, you can recover them by inputing the answers again, instead of doing the whole test again.
- It would be more adapted to professionals or large organisations.
- Some people have some emotional attachment to paper and pens, especially Sensing types.

I'm open to relevant suggestions. For example :

- unusual ways to assess types or physiological configurations.
- documentation on information elements, dichotomies, physiological configurations/properties or such.
- algorithms I can use to assess types.
- etc.

Come on, be bold and creative ! But please remain serious.

[bibliophile8]

I think something like rmcnew's picture questions would make a great section for the test. The combination of sensing imagery and intuitive concepts make it very effective in my opinion. Plus, they are less pointedly obvious.

[anndelise]

mcnew (or hugo, but i think it was mcnew) once had a test that had us read a written scenario (but i think even a video scenario would work too) and then questions were asked based on that scenario.

the biggest problem with it was that the answers to choose from weren't ...mmm.... real
as in they weren't really things people would pay attention to all-together.

since then i've imagined someone (else) creating a test that gives a scenario and then asks questions that would require people to give open-ended responses to answer it. (kind of similar to expat's recent attempt with the woman, the beer, and her motivations...only with better written questions)

Kind of like those school reading comprehension type tests, only designed to see how much of a certain type of details people get from a scenario...and what they then do with that information. (example, scenario is set up and expat's woman is receiving her beer, you notice that it is the wrong beer, what do you do? or she notices it's the wrong beer, what does she do? describe her clothes. describe the person sitting next to where she was standing. whatever's needed)

the scenario is given
the scenario disappears and cannot be backtracked to, nor copied/pasted
there are different scenarios and no two people will likely get the exact same ones (nor in the exact same order)
someone(s) would have to spend time deciphering the responses back
and people who are actually of the function/quadra/whatever that a question is being tested for should be part of the scenario creation and question creation process so that the questions/scenario can have info designed to draw from the actualy type/function combo and not some other types bias on what they think a different type/function combo should give.




ultimately, a video game type thing where you are interacting with characters and items and actually doing problem solving, etc. the game keeps track of which info metabolism pathways the player keeps referencing as the player goes through the game


however, i don't believe, that at this time, that socionics (especially this forum's members' understandings) is clear enough in being able to actually test for these things...particularly without trying to continue to turn socionics into a trait theory instead of the information metabolism and interaction theory it was intended to be. Too much subtleties get overlooked, lost, and flat out ignored on this forum. So I don't expect to see any really good tests come out for a few more years at least.

[marooned]

This is totally unhelpful, so I apologize beforehand.

I'm starting to wonder if in-person is really the best way to test people's types... I'm not sure there is any magic test that could do it all that effectively if not in person (though perhaps as "essay" test could work).

The other thing I was thinking of is have someone watch something, like a brief movie and film their face as they're watching it. Then have them talk about it after it's over. I've noticed that I pick up on certain elements in films and seem to ignore others, and everyone does this. But if you discuss something you watched with a lot of people, of course they all picked up on different things. Anyway, it doesn't need to be something like a movie, but it seems like it would work the best if it was in-person and spontaneous, whatever it is. That way you can get at how they're metabolizing information, without all the other crap getting in the way (where that crap is their self-image, their over-analyisis, their thorough thinking and introspection, their misconceptions, their mood of the moment, etc.). I can't really see a clear cut path to get around all of that stuff with an online test without making the test so bizarre that it doesn't really work very well... or works well for some types and not others...

This doesn't mean you shouldn't try though. Tests can always get better and yield more accurate results.

Unless you could employ an online vid and somehow try to work with that with carefully placed multiple choice questions... (just as a part of the test)...

[machintruc]

So far, we found those ways :

- pictures
- scenarios
- questions with open-ended responses
- problem solving-based typing
- movie typing

"video game" typing :

That's pretty classical. It reminds me some sort of program I created to test , which was based on counting objects either :

- In an unlimited, but counted time, trying to get the exact number of objects
- In a limited time, trying to get an approximate number of objects

Unfortunately, I don't have a copy of it anymore, but I can redo it. It may even be in an hardware way (i.e. counting physical objects)

Movie typing :

It's like Filatova's picture test : do you have descriptions or feelings ? (evaluates E/I) ; Do you describe/think of details or the big picture ? (evaluates S/N)

I think of another thing : which dude/girl do you resemble the most ? But it works better with physiological types than socionic types. I should find 27 non-copyrighted videos for that.

Open-ended answerable questions :

Though quite easy for a trained human to interpretate such answers, it's quite hard for software. I mean, does it have to do google-like searching ?

Scenarios :

Questions are basically based on parts of scenarios. But the problem is that you may not act like you describe in the scenario as you act on Real World. It may reveal self-image rather than cognitive/physiological patterns.

Pictures :

They should be taken with caution, because they may reveal self-image rather than cognitive/physiological patterns.

An SEI-4 dude/girl may prefer "heroic and chaotic" pictures because he's got an E8-like worldview, not being E8 him/herself though, whereas an LSI may prefer "peaceful and orderly" pictures, not necessarily because he's or E1, but because he's got a "conservative" self-image (he may be E3).

[guest32456]

I'm currently thinking of an easyTIM-like project.

I like your idea, Machintruc

Good luck!