I know that the meaning of words is determined by those who use them, but is there a specific number of people who have to agree on the definition of a word in order for it to appear in the dictionary? I ask because I'm having an online debate about the difference between atheism and agnosticism with several other people and I'm trying to explain to some of them that it's important that people have the same understanding of what those terms mean before engaging in a debate. When I pointed out to one guy that he didn't get to decide what all the words in the English language mean, I was accused of an ad hominem.
Usually, the meanings of the words are determined (by a lexicographer) by examining a corpus of text (usually some mixture of news, novels, and non-fictional texts).
However, in humanities and some fields of science things are different: There is often a specific meaning of a certain word assigned to it by a single author (and other others assign a different meaning to the same word), and one has to cross-check that one uses the word in the same meaning as the author one refers to uses it.
The primary source of dictionary definitions is authoritative dictionary definitions. The first cut would be between (online) lexical-redistribution systems which just take definitions from dictionaries and put them online, leaving information content unchanged. I ignore those. There are dictionaries that do some work on definitions; of these, the Oxford English Dictionary and the Merriam-Webster dictionary have a substantial research underpinning.
The general strategy is to not reinvent the wheel unnecessarily, so that definitions tend to be very similar across editions. There does not appear to be any automatic rule regarding what kind of attestation is sufficient vs. insufficient to warrant a change – furthermore, the driving goals of dictionaries are not constant (OED and Webster's both aspire to being descriptive rather than prescriptive, which leads to a volley of denunciations when an evil new word is added, such as "ain't"). The tendency is to base modifications in meaning to those motivated by "upper-echelon" sources. There is no attempt to sample usage by the masses.
Reading definitions in the OED (full version) is particularly informative, since they are quite happy to list all of the senses of a word and gives ample historical attestations, for which reason OED is my go-to authoritative dictionary. Almost no dictionary, though, gives really current definitions. The exception is Urbandictionary, but the problem there is that you get whatever (mis)understanding of a word a user feels like submitting, and I think there is a certain amount of vandalism (esp. lexicalized phrases that refer to shocking or repulsive situations).
You will not find any source that gives an exhaustive listing of all of the definitions of a word "out there", because there are too many hundreds of millions of people to survey. The approach that good dictionaries take is to focus on respectable sources, and in that sense they do not reflect contemporary usage.