For context I'm a 76 year old Boomer who took the SAT in 1965 and went on to earn a degree in Chemical Engineering, followed by an MBA and then much later in mid-career by a BSN.
1. The SAT is the best, and fairest, predictor of college performance because it is OBJECTIVE, and it should be made even more objective by eliminating any 'essa…
For context I'm a 76 year old Boomer who took the SAT in 1965 and went on to earn a degree in Chemical Engineering, followed by an MBA and then much later in mid-career by a BSN.
1. The SAT is the best, and fairest, predictor of college performance because it is OBJECTIVE, and it should be made even more objective by eliminating any 'essay' component from the Verbal section: Verbal REASONING skills (which are largely innate) need to be evaluated, not WRITING skills (which can largely be taught). So eliminate any essays entirely.
2. The "Adaptive Questions" system is actually quite good and efficient, and I have both stochastic comfort (being an engineer), as well as first hand experience with it:
After 20 years working as a Chemical Engineer I changed careers in my 40s and obtained a second Bachelors degree in Nursing in 1995 (BSN). That also happened to be the first year that the Nursing Licensing exam in New Jersey (NCLEX) went both digital and used "Adaptive Questions". It did NOT 'lower the bar'. To wit:
3. As I recall, back then there were a total of 255 potential questions of which a minimum of 75 had to be answered. The test would stop as soon as the algorithm was able to determine your knowledge level with 75 questions being the minimum neede to reach such a determination.
4. All of the top students in my class (myself included) received the minimum 75 questions and passed, implying high consistancy in answer quality. In fact all of the students in my class who had to answer less than 100 questions also passed. Above 100 questions and some classmates started failing, with the numbers who failed increasing almost linearly with the number of questions required, implying inconsistancy in answers (AKA 'guessing').
5. So the "Adaptive Questions" method makes it harder to get lucky by guesing the correct answer, whiich is easier to do with the 'total number of correct answers' method.
I agree that they need to be transparent with how the adaptive method generates a specific numerical score, but it can be done. For example:
With the NCLEX test that I took back in 1995 if you answered a question correctly the next question would be harder. If you answered incorrectly the next question would be easier. This pattern would repeat until you were getting consistently getting 50% of the questions correct and 50% wrong. The level of difficulty of the questions at which the 50-50 point was reached determined whether you passed or not. So to generate a specific numerical SAT score one might use the level of difficulty at which the 50-50 point was reached. With the present setup you would have to have 800 levels of difficulty which is a bit overly fine a graduation (not impossible, but not easy), - however you could certainly do it with 100 levels of difficulty and then the SDAT scores would be 1 to 100.
For context I'm a 76 year old Boomer who took the SAT in 1965 and went on to earn a degree in Chemical Engineering, followed by an MBA and then much later in mid-career by a BSN.
1. The SAT is the best, and fairest, predictor of college performance because it is OBJECTIVE, and it should be made even more objective by eliminating any 'essay' component from the Verbal section: Verbal REASONING skills (which are largely innate) need to be evaluated, not WRITING skills (which can largely be taught). So eliminate any essays entirely.
2. The "Adaptive Questions" system is actually quite good and efficient, and I have both stochastic comfort (being an engineer), as well as first hand experience with it:
After 20 years working as a Chemical Engineer I changed careers in my 40s and obtained a second Bachelors degree in Nursing in 1995 (BSN). That also happened to be the first year that the Nursing Licensing exam in New Jersey (NCLEX) went both digital and used "Adaptive Questions". It did NOT 'lower the bar'. To wit:
3. As I recall, back then there were a total of 255 potential questions of which a minimum of 75 had to be answered. The test would stop as soon as the algorithm was able to determine your knowledge level with 75 questions being the minimum neede to reach such a determination.
4. All of the top students in my class (myself included) received the minimum 75 questions and passed, implying high consistancy in answer quality. In fact all of the students in my class who had to answer less than 100 questions also passed. Above 100 questions and some classmates started failing, with the numbers who failed increasing almost linearly with the number of questions required, implying inconsistancy in answers (AKA 'guessing').
5. So the "Adaptive Questions" method makes it harder to get lucky by guesing the correct answer, whiich is easier to do with the 'total number of correct answers' method.
What you describe is great as a pass-fail test. But, the black-box of adaptive scoring isn't clear how it would give 400-800 scores.
I agree that they need to be transparent with how the adaptive method generates a specific numerical score, but it can be done. For example:
With the NCLEX test that I took back in 1995 if you answered a question correctly the next question would be harder. If you answered incorrectly the next question would be easier. This pattern would repeat until you were getting consistently getting 50% of the questions correct and 50% wrong. The level of difficulty of the questions at which the 50-50 point was reached determined whether you passed or not. So to generate a specific numerical SAT score one might use the level of difficulty at which the 50-50 point was reached. With the present setup you would have to have 800 levels of difficulty which is a bit overly fine a graduation (not impossible, but not easy), - however you could certainly do it with 100 levels of difficulty and then the SDAT scores would be 1 to 100.