Financial Mathematics Text

Sunday, December 2, 2012

Do philosophers know particulars?

In Methodists Vs Particularists, I suggested that philosophers who are particularists may not, in fact, be so. It's possible that there are no particulars that are known per se but rather a broad set of "methods" that are employed. This may not be explicit criterion per se; as a result I shall refer to these simply as "heuristics".

We'll consider a common "simple" example in philosophy: color identification. I believe that most children learn color identification via ostensive definitions. Consider the following scenario:

A parent might point to an object and say "red" and the child may repeat. The parent points to a second object and again says "red". The child repeats. The child may point to another object (which may or not be "red") and say "red". If the child has done so correctly the parent will give some positive encouragement. If the child has not, the parent might correct the child: "no, that's blue."

In the case where the parent transmits the claim that a particular object is "red", the child is gaining knowledge of a particular by means of authority. At this point the child has no idea how to make judgements about new items; the child merely knows that one particular item happens to be red and the child knows this because the parent has told the child as such.

But the child eventually learns how to identify new objects as either "red" or "not red". How is this knowledge obtained? While this is largely an empirical question (and therefore, not a "philosophical question"), I believe a conjecture can cause some doubt as to whether or not the child knows particulars. (To clear, the child knows the particulars that were told to it by the parents but that's not really going to address any "problem of the criterion" in any philosophical sense since testimony isn't considered terribly "grounding".)

Let's consider another scenario: Bayesian Spam filters.

The idea is pretty simple. Take some samples of "spam" emails and "legitimate" emails. Allow some statistical analysis to be done on the emails (e.g. look for differences in word or phrase frequency between spam and legitimate emails). Then given a new email and the words/phrases that occur within it, how likely is it that it is "spam" or "legitimate". Then some criterion is used, some threshold, to decide which category to place it in.

In this scenario, the initial emails given to the filter are the "known particulars" and these are known via some form of "testimony"; at the very least they are given. The probability analysis is a kind of criterion for making judgements. It is no doubt fallible but can be improved upon by correcting it as it makes errors in judgement. So the Bayesian Spam filter knows some particulars (those things it is told) but otherwise it is using a kind of criterion, one that is constantly adapting. This criteria is needed for the spam filter to be able to adequately handle new, previously never before seen, emails.

I would like to posit that the human brain does something similar when analyzing colors. As a result the claim that I know this particular ball (which I had never encountered before) is "red" is not exactly a "particular" I know but is based on a sort of "criterion" or "heuristic". It's not an explicit criterion and it may have to evolve over time but I think it shares similarity with utilizing a criterion.

If this conjecture on my part is roughly descriptive of how humans make judgement about, say, colors, then I have to conclude that it raises doubts as to whether or not human beings "know" particulars (without appealing to some sort of method or criterion).

This is not to say that one cannot question a particular method or criterion on the grounds that it fails to correctly identify something. But this may be more of a matter of privileging one criterion over another when the two methods diverge in their results.

Later on I hope to take a closer look at Gettier counterexamples as a case study.

No comments:

Post a Comment

Some common OpenID URLs (no change to URL required):