Shades of Grey and the Sorites Paradox

There's an interesting philosophical attitude being displayed in the Conspiracy Theory forum in a discussion about witness reliability. Here are the relevant posts:

Eckolaker said:

Would I agree people have a hard time recalling events? Hell no, anyone with any minor backround in psychology could tell you that the human brain is capable of recalling the most vivid of details and even more so when the event in question was extremely significant.

Click to expand...

CurtC said:

And that's why people who witness a crime always agree with each other 100%. Yes, I've been on three criminal juries, and I can state for sure that all the witnesses in each trial all told stories that didn't contradict each other at all.

Click to expand...

Horatius said:

Assuming you're being sarcastic here*, then they were obviously lying, as all the CTist will tell us that either lying or complete accuracy are our only options.

I'm growing quite fond of this black&white, black is white world they live in. It's ever so much easier than thinking, and more fun to boot!


*All I have to do now is figure out how to not figure out sarcasm, and the transformation will be complete. Then I can go take on R.Mackey!

Click to expand...

Eckolaker said:

There is no such thing as a Grey area in life. You either did or did not kill someone. You either do or do not drive. You either do or do not pay taxes. You either do or do not work. You either do or do not believe in god. (Even Jesus himself was quoted in the bible as saying this)

Those who talk about Grey areas are those that make excuses for their lives.

Click to expand...

maccy said:

If you believe in God and are a member of a Baptist church, do you believe in the same God as a Methodist or a Unitarian or a Mormon or a Jehovas Witness? How about a Catholic? A Greek Orthodox? A Jew? A Muslim? A Hindu?

If it's a clear case that there's this one thing, God, that you either believe in or don't, why aren't all these religions the same?

Furthermore, is it true that you are either short or tall? A child or an adult? An embryo or a human being? Where you do draw the line? What makes you correct in putting the line where you do? What is the significant change that occurs when that line is crossed.

If someone dies as a result of your actions does that mean you murdered them?

Why do you think criminal courts determine guilt "beyond reasonable doubt"? If the world is as black and white as you say it is, why not "beyond any doubt"? Why is the standard of proof in civil cases set "on the balance of probabilities?"

Here are some links you might want to look at:

http://plato.stanford.edu/entries/sorites-paradox/
http://www.btinternet.com/~justin.needle/
http://www.philosophers.co.uk/cafe/paradox4.htm
http://en.wikipedia.org/wiki/Paradox_of_the_heap

Click to expand...

So. what do you think? Is Eckolaker right about there being no such thing as a grey area?

No.

Yes. No. Maybe!


Definitely Maybe!

Of course, they are idiots and you pretty much established that by example. They seem to be stating that it is impossible for an eye witness account to be wrong, ever, unless they are lying. I fear if the day should ever come that I am left to the decision of a "jury of my peers" if this is the way they think.

If you could, try asking them if they have ever seen a magic show or an optical illusion of any sort. Ask if they've ever seen something at a glance and later found out they were mistaken as to what it was. Though, with that attitude I wonder if they just convince themselves they were absolutely right the first time and any account by someone else trying to correct them is a lie.

maccy said:

So. what do you think? Is Eckolaker right about there being no such thing as a grey area?

Click to expand...

Given the paucity of his explanation, and the counters you gave on continuum versus step function, he's like most CT's, full of crap.

DR

Reading the stuff about the sorites paradox, seems to me the easiest resolution is not to have two settings like a light switch but in fact implement a gradient like a dimmer dial. Things can be "more heaplike" or "less heaplike". In the case of baldness, no hair at all is the most bald, and the more you add the less bald that person is.

I have pubs on false memory-- in decent journals even. It's one of the few things I think I am trully qualified to speak on (for example, I have served as a peer reviewer for other people's work on FM).

False memories are trivially easy to demonstrate in a lab. The issue is how well do these results generalize to real world scenarios (like in eyewitness testimony)? My take on it is that specific data points and base rates for false memory in the lab likely do not generalize to the real world (that's not the point of the research, anyway), but the cognitive processes that underlie false memory in the lab are the same as those that produce false memory in the real world.

So, I wouldn't put much weight in eyewitness testimony alone-- especially for the specifics of the event (specifics could even be what the suspect looked like, or mistaking the suspect's face for some other vaguely familiar face). With other hard evidence, eyewitness testimony could be informative; without it, good luck!

Another criticism raised in lab research is that the false memories produced there are not for emotionally charged events (whereas witnessing a rape, or crime, or having a relative abuse you is highly emotional). Critics argue that when the event is vivid and emotionally charged, we should be less likely to commit false memory.

The best work I've done (so far, I hope!) is a paper showing that people can falsely remember seeing obscene words (f*ck, *****, bitch, penis, etc) after studying a list of words that contained no obscene words. I acheived this by making the study list be words that were orthographically related to a target swear word (the swear word iteself not appearing in the study list).

For example, having words like witch, ditch, butch and botch leads about 20% of participants to misremember having seen the word bitch.

I had some manipulations that pushed this baseline rate up or down, depending, but the point is that false memory seems to extend to things beyond just mundane words (like chair or sleep).

It's known as the dirty word study-- Elizabeth Loftus cited it once (I also met her briefly, and when I told her of my paper, she said: Oh, you did the dirty word study, and then she shook my hand!).

bpesta22 said:

For example, having words like witch, ditch, butch and botch leads about 20% of participants to misremember having seen the word bitch.

Click to expand...

It would be interesting to design an experiment to try to determine if the remembered word later was a mis-remembering, or a proper remembering of a mis-read word. I'm sure a number will read "botch" and misread it as "bitch", as opposed to those who read "botch" and conflated it with ditch into bitch.

Beerina said:

It would be interesting to design an experiment to try to determine if the remembered word later was a mis-remembering, or a proper remembering of a mis-read word. I'm sure a number will read "botch" and misread it as "bitch", as opposed to those who read "botch" and conflated it with ditch into bitch.

Click to expand...

It would, but I suspect -- pardon me -- that it would be a bitch to design. And easy to botch.

Controlling for word frequencies is difficult. And finding the point where a mis-reading turns into a mis-remembring would be tricky. I suppose one possibility would be to give people a list of words and ask them to press a button for the vowel in each word. So if someone presses an "i" when they see the word "botch," it's a mis-reading, but if they press an "o," then it's a mis-remembering. But what do you do about "ditch"/"bitch"? Maybe present it as a test of lexical reproduction and the subjects have to retype the word they see (or type it backwards, or something stupid like that)?

I dunno. Glad I don't have to desigh and run these types of experiments any more..... RHIP and all that....

drkitten said:

It would, but I suspect -- pardon me -- that it would be a bitch to design. And easy to botch.

Controlling for word frequencies is difficult. And finding the point where a mis-reading turns into a mis-remembring would be tricky. I suppose one possibility would be to give people a list of words and ask them to press a button for the vowel in each word. So if someone presses an "i" when they see the word "botch," it's a mis-reading, but if they press an "o," then it's a mis-remembering. But what do you do about "ditch"/"bitch"? Maybe present it as a test of lexical reproduction and the subjects have to retype the word they see (or type it backwards, or something stupid like that)?

I dunno. Glad I don't have to desigh and run these types of experiments any more..... RHIP and all that....

Click to expand...

It was a recognition test, and I did have control lists (non emotional words) that were matched to a swear word on things like word frequency.

Here's my study lists!

bpesta22 said:

It was a recognition test, and I did have control lists (non emotional words) that were matched to a swear word on things like word frequency.

Here's my study lists!

Click to expand...

Oh, I didn't mean to imply that you weren't aware of word frequency issues; I was simply commenting to Beerina that psycholinguistic experiments are brittle -- if you try to bend these kind of studies too far, they break.

bpesta,
What do the (numbers) signify?

kmortis said:

bpesta,
What do the (numbers) signify?

Click to expand...

Word frequency-- number of times you'd see the word per million words

So, I'd expect to see "rink" as often as "whore"? Ok. Thankee.

kmortis said:

So, I'd expect to see "rink" as often as "whore"? Ok. Thankee.

Click to expand...

Yes, at least in 1968, which is the most recent set of norms available!

As you might imagine, it's a massive pain in the ass to do word freq. norms

bpesta22 said:

Yes, at least in 1968, which is the most recent set of norms available!

As you might imagine, it's a massive pain in the ass to do word freq. norms

Click to expand...

Noooooo. You mean you can't just record the speech and writings of everyone, everywhere and count them up? What's all this technology good for anyway?

Seriously though, sounds like some neat research. I remember Mind Hacks doing something like this. It's been over a year since I read the book, so I don't remember what the lists were for, but they had word lists.

bpesta22 said:

Yes, at least in 1968, which is the most recent set of norms available!

Click to expand...

Surely you jest. You're still using the Brown corpus? What's wrong with the British National Corpus norms?

Development of large corpora is something of a cottage industry and has been for the past fifteen years.

drkitten said:

Surely you jest. You're still using the Brown corpus? What's wrong with the British National Corpus norms?

Development of large corpora is something of a cottage industry and has been for the past fifteen years.

Click to expand...

Kucera Francis.

I collected the data about 6 years ago-- that was the most recent one I was aware of. I still see americans citing K&F today.

Got a quick link to BNC?!

Word frequency wasn't a horrible problem in this study as I rotated which study lists the subjects saw (half got set A at study; the other half, set B). So, I could look at baseline false alarm rates to the words when people did or did not study the orthographically related list.

drkitten said:

It would, but I suspect -- pardon me -- that it would be a bitch to design. And easy to botch.

Click to expand...

Okay, now you're having too much fun.

bpesta22 said:

Kucera Francis.

Click to expand...

Yeah, that's the Brown corpus (so-called because it was collected at Brown U.)

Got a quick link to BNC?!

Click to expand...

Try here.

There's a lot of other, newer, stuff that's even larger than the BNC -- but at least the BNC is only ten years out of date, not thirty. More importantly, the BNC is a hundred times larger than the Brown corpus, so the data are that much more accurate. And most of the larger corpora than the BNC (I think Turney-Littmann used a hundred billion word corpus) have balance issues (e.g. they're all from Web pages or something.)