I clearly don't have all the answers, and you raise many legitimate points. I still think "grades" as we consider them today function to continue to degrade education quality in the united states.
Goodhart's law would be better written as "when a proxy measure becomes a target, it ceases to be a good measure", which both gets to the point but also helps consider that sometimes proxies are the best we've got. Sometimes one just needs to measure better.
Yes, and grades are functioning as a proxy for far too many things to be a good measure of nearly any of them at this point.
Have we? Or have we allowed grades to get diluted due to perverse incentives? Is the appropriate response to surrender on that or to attempt to realign via better tests? It's pretty important to figure out what the foundational demand is or else we'll just end up reinventing the wheel with new labels. Like with grades, yes the true goal is a complex combination of "how can we motivate students to learn, and measure how well they're learning in a dynamically granular manner, and also measure how well they handle meta tasks related to learning that matter for real world unsupervised activities including continued learning and learning application, and then also convey all that in a high density form that decision makers in different domains can make use of before the student in question has more track record, and also we need to feed that back into how we ourselves are doing with our teaching/methods". That's not an easy problem space, but it's key to recognize that grades became and have continued to be a thing because they've been useful.
Here you point out that grades have a crazy number of competing uses that create perverse incentives and dilute their meaning. Despite these things they've been a simple, intuitive, and useful enough proxy to reach the point they're at today. My argument is that they have been becoming less useful over time as more and more people have the capacity and incentive to game them. LLM's are, I believe, the tipping point where they no longer do what we need them to do well enough to keep using them.
But students very much actively want to be able to convey their achievements outside of the class. The whole point for most students is linked to the real world, to getting good jobs or other opportunities they desire. "Consequences" are therefore fundamental, touching on various zero-sum games. Of course people are going to care, and some people are going to do better relative to others, and the ones who do worse are going to have incentives to cheat. You can't just side step all that. Like if some colleges stop providing grades as you suggest, game theory out a bit the results of that. Students exit college, and go to apply to some job, and naturally the company wants to know how they did. How do they convey this in a rapid objective fashion? Will companies then preferentially recruit from colleges that do still provide grades? What would the result when it comes to where future students apply to, and the consequences for the colleges without grades they don't? Etc.
Not having the consequence be a direct impact on opportunities and learning potential is not the same as there being no consequences at all. It's about re-aligning the consequences to better match actual learning/effort, and then finding a different (related) measure that better matches capacity for success.
I'm not saying there should be no way to differentiate relevant student skills and achievements, just that grades in their current form are an increasingly bad proxy. If the solution were clear we wouldn't be having this debate. Any alternative would need to be trialed, tested, and refined tremendously, and there hasn't been the will/sufficient incentive to find an alternative to grades yet. Perhaps we're there now thought. Maybe we'll default to more rigorous high stakes proctored testing like the SAT. Perhaps we'll need to fall into more subjective qualitative assessments from teachers throughout schooling that are then processed into standardized themes by a specially trained LLM. The lack of an obvious alternative should not prevent us from noticing that we need to really start looking for one.
Teachers need some form of grade within a class to track individual understanding and development in order to adequately lesson plan and target interventions, and ultimately in order to do a good job teaching. When we use the same grades outside of class as a proxy for potential for success, we create an incentive to interfere with the accuracy of grades as a teaching tool and consequently decrease their value overall. That used to be tolerable when a small group of wealthy folks were the ones capable of effectively exploiting this issue. LLM's offering that capacity to everyone adds just too much bad data to the mix. Garbage in, garbage out, and LLM's introduce too much error for grades to be a meaningful measure now.
