cool work! in the pinterest work, you feed extract features from posts conveying methods/instruments for math instruction and use them to categorize based on bloom's hierarchy of cognitive demands. I wonder whether this project should incorporate the quality of the method/instrument either as a feature or even an outcome measure? this issue might be avoided with your experts doing feature extraction, but i'm sure it's inevitably the case that a pin has a consensus assignment to some hierarchy level, yet it's still a relatively bad tool for education at that level. This confound of quality could disrupt your model and/or make its predictions less useful to educators. Is this something you've considered?
cool work! in the pinterest work, you feed extract features from posts conveying methods/instruments for math instruction and use them to categorize based on bloom's hierarchy of cognitive demands. I wonder whether this project should incorporate the quality of the method/instrument either as a feature or even an outcome measure? this issue might be avoided with your experts doing feature extraction, but i'm sure it's inevitably the case that a pin has a consensus assignment to some hierarchy level, yet it's still a relatively bad tool for education at that level. This confound of quality could disrupt your model and/or make its predictions less useful to educators. Is this something you've considered?