Hello Christian,

The original attempts at adaptive tests (e.g., self-scoring flexilevel test, stradaptive test) simply took a fixed item pool, sorted them by difficulty, and started an examinee in the middle. Successful attempts (correct answers) routed to the next more difficult item; unsuccessful attempts routed to the next easier item, up to a fixed number of items (rather than having the standard error of ability estimate fall below some threshold). These were initially offered up in paper-and-pencil form, but could easily be computer-administered as well.

In the 1960s-1980s, there was a glut of interest in patient management problems (for assessing the diagnostic/treatment skills of preservice medical students). These also were adaptive, in that one's choices at a stage governed the subsequent information and question challenges that were presented. Again, the bank was fixed, the routes were fixed, and scoring was often based on whether a response was productive, unproductive (but benign), or harmful.

There was also another method, developed by Sato, called the Subject-Problem (S-P) method, which was much like IRT scaling, except essentially involving ranking of persons (based on total scores) and items (based on rates of correctness). In a sense, a "perfect" item set would behave like a Guttman scale.

Is this what you're asking about?

It's fair to say that, with the wide-spread use of IRT models, that a lot of the hurdles that previous attempts encountered were ameliorated considerably. Not the least challenge of which being how to add items to a pool and not distort the meaning of the (original) score scale.

Good luck with your work.

Thank you! I didn't expect as much! You made my day!

This essay is about a new theory called General Total Score:

Preprint Item Response Theory and Its General Total Score