> If faculty bonusesOy. Faculty don't get meaningful bonuses. Like, a $100 gift card to the university book store is a really nice gesture for someone who really went above and beyond. True story.
Not only do they not get bonuses, but even their base salaries are often 1/2 or less of base salaries in industry. Meaning total comp can be as low as 1/3rd or even 1/5th for competitive applicants (100k vs. 300k is a pretty common spread in CS. The spread gets even larger for top-tier candidates because academia has a hard ceiling but industry is more than happy to throw 400+ and top-tier fresh phds...).
>... included a mix of pass rates, success in subsequent classes, AND textbook cost, this problem would go away.
It consistently amazes me that people underestimate the difficulty of perfectly projecting a high-dimensional approximation of an inherently unknowable quantity onto a single scalar.
If it were actually possible to do what you prescribe, to reduce good teaching down to a measurable combination of measurable variables, then why on earth wouldn't we just replace all teaching with a reinforcement learning algorithm?
Oh, and BTW, things like MyMathLab aren't even saving professors' time. They are saving graduate TA time. Those TA's are making 20k-30k a year.
The reason that auto-grading software can charge a lot of money is that even horrendously underpaid graduate TA labor is extremely expensive compared to a software license.
So, I think subjecting faculty to teaching metrics is probably a terrible idea, especially in STEM at lower tier universities. Recruiting is already hard enough. Doing so is not going to have the intended effect and might even cause a brain drain.
Projects like this pilot are a much better idea. We should also have pilot programs for open-sourced auto-grading software and small NSF grants (e.g., $8k over 3 months -- a princely some for a grad TA!) to cover the cost of refreshing the question banks every few years.