Everything is uncertain, including uncertainty, for the simple reason that your brain is just a machine, and machines can fail for a variety of reasons. If you're willing to accept that something as simple as 2+2=4 is slightly uncertain (and you should be) then you should be willing to accept that the reason 2+2=4 is uncertain is also uncertain. Consider that "the reason 2+2=4 is uncertain" is significantly more complex reasoning than "2+2=4" itself, and more complex means more error-prone.
Anyway, why is everything uncertain? Because your brain is a machine, and machines are fallible. Let's start by looking at how silicon computers are fallible. Silicon computers are much more reliable at accurate math calculation than your brain is, and yet you cannot fully trust the result of a mathematical calculation done on a computer.
- There can be software bugs causing inaccurate calculations. The types of bugs are too numerous to list, but can include failure to bounds check, failure to initialize, floating point overflows, rounding errors, prematurely freed memory, and on and on and on.
- Someone could have hacked your system without you knowing, and maliciously altered your program to give you the wrong calculation result.
- There can be hardware bugs causing mathematical failures, such as https://en.wikipedia.org/wiki/Pentium_FDIV_bug or others we may not know about.
- Even if there are no hardware design bugs, there is always a chance of manufacturing defects, e.g. a stray bit of dust on the lens when engraving the wafer. The manufacturer tests for such defects but no testing process can give a 100% guarantee.
- Cosmic rays or other radiation could have altered the functioning of the chip at a critical time
- You may have undetected age-related failure in your computer's memory or hard disk
- It's extremely unlikely, to the point that it's probably never happened in the history of computing, but there's still a nonzero probability that quantum tunneling of the electrons on your chip could alter the result of a calculation.
So those are some examples of how a machine can fail at any calculation. Your own brain is a machine too, and a much less precisely designed one than a computer. Your brain makes mistakes at a much greater rate than a computer.
- Everyone is familiar with making arithmetic and logic mistakes on academic tests. Even when someone fully understands a subject, they still make mistakes due to inattention or dyslexia/dyscalculia. And people are not fully aware of when they are inattentive; that's another mistake we make.
- Neurons operate probabilistically; action potential produces a chance of firing, not a certainty. This means that the result of any calculation done by neurons is at least partially random, and could randomly give some very incorrect result.
- Cosmic rays have a chance to affect neurons just as they can affect computer chips
- Neurons die, and sometimes are born. What if that happens in the middle of the calculation you're performing?
- Artificial neural networks are trained in a process of trial-and-error, and the training may give good results but it does not result in a network with guaranteed perfect performance on a problem. If your brain's neurons are anything like that, they also do not have any guaranteed perfect performance.
Of course, if your brain's operation is fallible, then any calculation or reasoning you do mentally, no matter how simple, might in fact be an error, as it is the result of your brain's operation. You have no way of verifying or checking this reasoning except by ultimately relying on your brain, so any verifying/checking process is also fallible. Perhaps you can get very close to certainty, but it is irrational to suppose you ever have perfect certainty.