Since scientific laws are approximations or observed regularities and generalisations(or an attempt at finding the actual laws of nature) rather than the actual laws of nature -- does such a description assume that laws of nature still do exist ? If not, then would a world with no absolute, immutable and eternal laws be chaotic ? Would science still work without absolute laws, or is it necessary for absolute laws to exist for science ?
It is true that all physical laws are convenient and useful approximations valid for everyday use within known ranges of applicability.
However, without immutable physical laws i.e., that do not shift about randomly over time, the universe would indeed be chaotic. Energy would not only not be conserved, it would be mathematically meaningless, as would measures of distance and time. Material objects could pop into and out of existence at any time, one year could equal one second or one million years at any time, and so on. In fact, a universe without physical laws that do not change would not furnish the opportunity for life to evolve in it- so we wouldn't even be here to observe the resulting chaos in the first place.
And no humans would mean no science.
In Quentin Meillassoux's After Finitude, he makes some very interesting points about this idea.
In the chapter called "Hume's problem", he gives a fairly lenghty and sophisticated refutation of the following Kantian argument :
(1)If the laws of nature could change without any reason -i.e if they were not necessary- they would change frequently.
(2)The laws of nature are not changing without any reason (they aren't changing at all, really)
(3)Consequently, the laws of nature cannot change without any reason, i.e they are necessary.
Basically, Meillassoux highlights the crucial ontological presuposition of the "frequential implication" in 2, and shows (using Badiou's Being and Event) that it is flawed to apply this kind of probabilistic reasoning to our Universe(taken as the "totality of what is possible").
Indeed, how could we know, for example, that the laws wouldn't change extremely rarely or extremely unimportantly, going as far as to destroy the possibility of science or existence ?
Science is the endeavour of creating models of reality that can be used to make predictions (and not only explain what happens, this is what makes the difference between science and pseudo science).
As such, it has to assume its subject of study behaves in a consistent, immutable manner. If the universe starts acting differently between the creation of your model and the moment you use it, you obviously can't make any prediction. When an engineer starts designing a wing profile for a new plane, they have to assume, implicitly, that material science and fluid mechanics are as valid as when they learned about it.
It can be a real problem for social sciences, where society changes as you try to find patterns of social behavior. Furthermore if the publication of your research leads to those changes...
So far, in 6000 years of recorded history and about as much of recorded research, it has been the case that the behavior of the universe does not change. I.e. models that have been subverted were because of new observations unavailable to previous research, not because things stopped working the way they were. The study of older materials like stones, fossils, radioactive decay or astronomic objects are consistent, and a solid hint that it has been the case since long before the dawn of humanity.
So it's a reasonable assumption to think the universe behaves in a constant, steady way. At least if it changes, the change is slow enough that we can't perceive it to the extend of our technology. Nothing guarantees it will always be the case, but it has not been observed so far.
If it does indeed change, but the change is gradual enough that we only need to slightly update our models every 10.000 years, science still has a role to play to build models valid for many generation, or even build a model of the change itself for extended validity.
Even if the change is irregular, like a sudden change with very long periods of calm, science still has a role to play in rebuilding our models from the ground. The only change pattern that would make science useless is if the deprecation of models occurred too often or fast for the endeavour of establishing them to be viable economically. (Why spend 3 years figuring how it works of it starts working differently in 2 weeks ?)
So science can afford a bit of chaos, if it was the case that the universe is indeed chaotic, and it would still work to a certain extent.