In general, they consist of some input, some internal process, and output.
For a large class of systems, a desireable feature is that they be in some equilibrium, most generally some dynamical equilibrium (static equilibrium is often boring) and possibly exhibiting some quasi-oscillatory, or limit-cycle, wandering away from the equlibrium.
For a stable dynamical equilibrium we generally want the departure from the mean equilibrium to be bounded, possibly strictly bounded, and most certainly not to exhibit, say, secular exponential growth.
Now, it is possible to do this by imposing some external constraints on the system, but ideally you want systems that are internally stable, purely by virtue of their own internal dynamical processes.
As we all know, to do that, in general, you need feedback!
Running complex systems open loop is an invitation to disaster, the system will make secular excursions from equlibrium and eventually run away, either to chaos or some other, possibly less desireable, equlibrium.
But, not just any feedback. If all the feedback is positive, then the system is almost certain to run away, and rapidly.
If you're lucky, a well put together system will fail safe, some governating assembly will either bound the excursion or terminate the processes leading to bad behaviour.
If not, then either the system breaks, or more commonly some weak point in the system breaks. You can even design these into the system, just in case you overlooked some feedback loop.
But, what you really need is negative feedback.
Negative feedback is kinda boring, it even sounds kinda boring, but it is what keep complex systems stable. Without it they are destroyed.
So, we must embrace negative feedback, we should wallow in it, we should absorb it, digest it, learn our lesson, and go on to absorb some new input to be processed into our wonderful output.
But, that is hard work, and rather tiresome.
Ah, well, don't know what made me say that.
Back to referee reports for me.