I'm an astrophysicist. I love the idea of shaking up how we do science, but MOND models really aren't that compelling for a practicing scientist (I would love to be wrong). I periodically consider working a modified gravity project, but it never seems that interesting. Maybe I can lay out my personal impressions for you in a software analogy.
There are simple situations where physics can be reasonably distinguished from the noise. Think Newton watching an apple fall from a tree -- these are test harnesses with a debugger and careful control of the environment. There are also astrophysical situations where it's very hard to squeeze insight out of the chaos. Instead of looking at an apple to derive Newtonian gravity, imagine instead trying to figure gravity out by looking at a flock of birds. There are tricky things in the way of our understanding, like lift, turbulence, and biomechanics. We hope to understand the bad situations someday, but right now it's tough. Instead of the nice test harness, these situations are like debugging through cryptic, un-reproducible user complaints. What's their OS? Do they have the right drivers installed? Do they have all of their ports blocked for some reason?
In astrophysics, one nice system is the Universe at very large scales: echos of the big bang, the clustering of matter, and the formation of structure. We have great data these days about the large scales, and general relativity (GR) + dark matter (DM) is a very predictive model here. To our chagrin, the data always matches the theory. It takes 5 or 6 parameters, but the model has withstood huge improvements in data quality without really changing since the 90s. The other nice situation is the small scales like our own solar system: we can measure things extremely precisely at home, and again GR works remarkably well. There are a few other good situations involving objects like pulsars. Finally there are the hard situations: intermediate scales that involve the messy physics of star and galaxy formation. We don't really know how these processes work, but we can cobble together simple models with DM that sort of match the data.
Modified gravities in the literature always seem to act exactly like general relativity + dark matter in the clean, understandable situations, although it takes various contortions (screening) for these models to do this. Modified gravity models seem to only act differently in the bad arenas that are hard to understand, when there trickier issues like galaxy and star formation. That's deeply suspicious, and really weakens the value proposition of these alternatives. It's like introducing a software feature to fix something that never happens on the dev machine, based on mysterious user reports that you can only half decipher.
In fairness the fine-tuning / epistemology / scientific method tap dancing seems to affect the string/supersymmetry crowd a lot more than the astrophysics crowd. The crossover act seems to be superpartners as WIMP candidates. My original comment didn’t reflect that.
Five or six freely chosen parameters is a lot. You know the von Neumann quote. If those parameters have continued to fit vastly more precise measurements without changing for 30 years that’s starting to sound like failure to falsify.
And if so, ultimately you folks have posed the particle/high-energy folks a problem: we found a shitload of weakly/non-interacting mass, where’s my particle bruh?
To the layperson, exotic undetectable mass sounds like something that will turn out to be a good metaphor for something deeper, but that doesn’t mean a lot via the definition of layperson.
Even laypeople know the string folks are full of shit: Witten has a Fields and not a Nobel for a reason.
There are simple situations where physics can be reasonably distinguished from the noise. Think Newton watching an apple fall from a tree -- these are test harnesses with a debugger and careful control of the environment. There are also astrophysical situations where it's very hard to squeeze insight out of the chaos. Instead of looking at an apple to derive Newtonian gravity, imagine instead trying to figure gravity out by looking at a flock of birds. There are tricky things in the way of our understanding, like lift, turbulence, and biomechanics. We hope to understand the bad situations someday, but right now it's tough. Instead of the nice test harness, these situations are like debugging through cryptic, un-reproducible user complaints. What's their OS? Do they have the right drivers installed? Do they have all of their ports blocked for some reason?
In astrophysics, one nice system is the Universe at very large scales: echos of the big bang, the clustering of matter, and the formation of structure. We have great data these days about the large scales, and general relativity (GR) + dark matter (DM) is a very predictive model here. To our chagrin, the data always matches the theory. It takes 5 or 6 parameters, but the model has withstood huge improvements in data quality without really changing since the 90s. The other nice situation is the small scales like our own solar system: we can measure things extremely precisely at home, and again GR works remarkably well. There are a few other good situations involving objects like pulsars. Finally there are the hard situations: intermediate scales that involve the messy physics of star and galaxy formation. We don't really know how these processes work, but we can cobble together simple models with DM that sort of match the data.
Modified gravities in the literature always seem to act exactly like general relativity + dark matter in the clean, understandable situations, although it takes various contortions (screening) for these models to do this. Modified gravity models seem to only act differently in the bad arenas that are hard to understand, when there trickier issues like galaxy and star formation. That's deeply suspicious, and really weakens the value proposition of these alternatives. It's like introducing a software feature to fix something that never happens on the dev machine, based on mysterious user reports that you can only half decipher.