Very Well Researched and written.Your analysis points out the outcome pitfalls that IB and other educational institutions should consider and correct for.
Here are some of the additional views to consider, based on my sense of data flow thru your model:
- the fact that your model can predict gender, race and socioeconomic status from the historical grades is rightly pointing out that today grades are indeed strongly (and very unfortunately) correlated with these extra-curricular environmental variables. This is a well established fact and many social programs have been designed in an effort to flatten this bias.
- so unless the model takes its own predictions and loops it back as X variables, it should not reinforce the bias. The model accurately predicts along the pattern of social bias that is Actually present today.
- If we are bold enough we can decide thru this model to introduce counter biases (feedback loop) against these factors so that we level the playing field and truly measure the raw intellect and not the Environmentally conditioned intellect - but that is debate for another time!
All in all a great research and write up - wake up call to folks who Blindly assume predictions for truth! I will wait for your next part - “how do we correct for these biases”!
Harsh Vinayak
Fantastic effort in asking the question and raising awareness. Hopefully the universities would consider the shortcomings of the process for the IB results to base their decisions.
Loved the animated GIF of "the model" literally falling!
The coronavirus has thrown a lot of curveballs and each has sent impacted groups looking for answers. As OP correctly concludes, there are no good answers. Only worse and bad ones.
Besides the original problem, the OP's analysis is well presented and brings out the worse in this not so well thought out solution. I hope IB is listening and is willing to adjust their model to address issues pointed out by the OP.
Thank you - I appreciate it. Some people on another forum actually informed me of the fact that other educational boards in the UK are planning on adopting this exact same 'model based assessment methodology'. I do not understand why more people aren't upset about this.
Use of statistical model to allocate grades to students may not be fair to many students. Models should be used for estimation and not actual allocation of grades.
Here are some of the additional views to consider, based on my sense of data flow thru your model:
- the fact that your model can predict gender, race and socioeconomic status from the historical grades is rightly pointing out that today grades are indeed strongly (and very unfortunately) correlated with these extra-curricular environmental variables. This is a well established fact and many social programs have been designed in an effort to flatten this bias.
- so unless the model takes its own predictions and loops it back as X variables, it should not reinforce the bias. The model accurately predicts along the pattern of social bias that is Actually present today.
- If we are bold enough we can decide thru this model to introduce counter biases (feedback loop) against these factors so that we level the playing field and truly measure the raw intellect and not the Environmentally conditioned intellect - but that is debate for another time!
All in all a great research and write up - wake up call to folks who Blindly assume predictions for truth! I will wait for your next part - “how do we correct for these biases”! Harsh Vinayak