Gettiers in software engineering (2019)

It's likely that I'm wrong, I need to look deeper into it.

But isn't the paper-mache cow case solved by simply adding that the evidence for the justification also needs to be true?

The definition already requires the belief to be true, that's a whole other rabbit hole, but assuming that's valid, it's rather obvious that if your justification is based on false evidence then it is not justified, if it's true by dumb luck of course it doesn't count as knowing it.

EDIT: Okay I see how it gets complicated... The evidence in this case is "I see something that looks like a cow", which I guess is not false evidence? Should your interpretation of the evidence be correct? Should we include into the definition that the justification cannot be based on false assumptions (existing false beliefs)? I can see how this would lead to more papers.

EDIT: I have read the paper and it didn't really change my view of the problem. I think Gettier is just using a sense of "justified" that is somewhat colloquial and ill defined. To me a proposition is not justified if it is derived from false propositions. This kind of solves the whole issue, doesn't it?

To Gettier it is more fuzzy, something like having reasonably sufficient evidence, even if it is false in the end. More like "we wouldn't blame him for being wrong about that, from his point of view it was reasonable to believe that".

I understand that making claims of the absolute truthfulness of things makes the definition rather useless, we always operate on incomplete evidence, then we can never know that we know anything (ah deja vu). But Gettier is not disputing the part of the definition that claims that the belief needs to be true to be known.

EDIT: Maybe the only useful definition is that know = believe, but in speech you tend to use "he knows P" to hint that you also believe P. No matter the justification or truthfulness.

EDIT: I guess that's the whole point that Gettier was trying to make: that all accepted definitions at the time were ill-defined, incomplete and rather meaningless, and that we should look at it closer. It's all quite a basic discussion on semantics. The paper is more flamebait (I did bite) than a breakthrough, but it is a valid point.

Indeed, there needs to be some sort of connection between the truth and the justification, not just "true && justified".

The problem is that when you're working at such a low level as trying to define what it means to know something, even simple inferences become hellishly complicated. It's like trying to bootstrap a web app in assembly.