There is an
argument that we can never
truly know anything. All our
sense experience is ultimately worthless, because it might be totally
deceived in some way. This is a common argument in
undergraduate epistemology, but is
invalid.
Epicurus dismissed it long ago. Here is my own
refutation.
The argument for undetectable illusion is contentless. It is logically the same as twitching as you say a sentence. It has no bearing on truth values. In this case it is not only a pleasure but a duty to use Ockham's Razor.
By undetectable I mean that by assumption no evidence can ever be found to unmask the "illusion" and distinguish the truth. The "illusory" situation is identical in phenomenology to the real situation, except that the reality is claimed to have some metaphysical property of "reality" which we by definition can never detect.
Let us take first an absolutely undetectable illusion, say that the universe came into existence with all its fossils and memories five minutes ago. How is this different from what we actually believe of its age? It is not. No sliver of distinction can be found between the two descriptions. They describe identical situations. All effects are identical. No sense can be made of the expression "five minutes ago", no physical meaning can be given to the concept of time, in the non-standard description. It is like drawing the timeline logarithmically instead of linearly. It looks different but it doesn't correspond to any factual difference.
So take solipsistic undetectable illusions. I suspect a perfect brain-in-vat simulation is algorithmically impossible, but let's assume it can be done. Then it is possible that you or I could be a brain in a vat, and that the rest of humankind or whoever is doing this to us may be aware of the fact: it is to them quite objective that a certain vat and a certain brain have certain observable inputs and outputs, and they can read off from them that an illusion of 20th-century Earth is being presented to the brain.
Note that the situation is not logically affected by the number of people being deluded. Two brains in two vats, connected to each other or not, or a whole community of brains, or a bottle planet of Kandor effect where all astronomy is an illusion created by extraplanetary beings. The imagined situation is just one where one group can objectively see the truth (it's only a giant planetarium or vat) and the other group are subjectively deluded into believing in something that's not really there.
It is this situation the logic of which I am now going to attack.
Remember that I am conceding that it is physically (and thus logically) possible. What I am maintaining is that that sort of "possibility" is of no interest, and is trivial and contentless, and is a misuse of the term 'possible'.
If one bore in mind the possibility of being a brain in a vat one might be tempted to allow for it in descriptive statements. So instead of saying 'This book is green' one might think it more accurate to say 'This book is green, unless I am a brain in a vat (in which case I don't know whether such a book exists)'.
Here are some more possibilities: the CIA has drugged me with a new drug that makes me confuse green and blue; a genie is following me around casting illusory books in my path; the book is blue but quantum fluctuations red-shifted the light travelling to my eye; I made a slip of the tongue; and so on. Infinitely many such invalidating clauses can be tacked onto any observation. The unless-clause in "The book is green unless..." can be arbitrarily convoluted. And as the same clause appears on absolutely everything, regardless of meaning and evidence and plausibility, it is meaningless. It tells you nothing. It adds no content.
Contrast this with genuine use of an unless-clause. "Mary's car is green unless...". What can factually go here? Unless she's bought a new one since I last saw it. Unless it's been spraypainted. Unless it's been destroyed in a crash, or she's sold it and hasn't got one. These are genuine ways in which Mary's car can fail to be green. They are non-trivial.
"Mary's car is green unless it's some other colour" is trivial and meaningless. It doesn't tell you anything about the car or the colour. But a very similar utterance might have content. Suppose Mary thought the green ugly and was thinking of having it spraypainted but didn't know anyone who did spraypainting. Or suppose Mary has always bought green cars but recently laughed at herself and said she ought to break the silly habit. In either of these cases "Mary's car is green unless it's some other colour" is both true and informative if said by someone who knew the context but hadn't seen the car today. It is slightly odd and would require clarification, but what statements don't?
In real language, unless-clauses alter the meaning and truth-conditions of observations. But where U is an undetectable illusion, "This book is green unless U" has literally and precisely the identical meaning to "This book is green". It is not more accurate, nor more cautious, nor less subjective. It is like "This book is green, touch wood" and "This book is green (<hesitant twitch>)".
Postulated undetectable illusions cannot affect meaning or truth, cannot enter into judgements or descriptions or observations. This is a logical fact, not an empirical limitation.
A description of the world in ordinary terms is identical in factual content with one that claims to take spurious unobservable possibilities into account.