Why do "unlikely" chance events seem meaningful?

[Sledge]

Ah yes, the master cynic at work. Perhaps it is you especially that should have a listen to the podcast.

Why? Is there something you would like to discuss about it? Your OP consists of "here's a podcast, anyone who's a real skeptic should listen to it." You've put me off on two grounds:
1) "True skeptic." So anyone who doesn't listen to this isn't a true skeptic. Who put you in charge of skepticism so you can make that determination?
2) Your ability to judge evidence is up there with Stevie Wonder's ability to judge paint colour. I'm going to need more than "I think it's good" to make me want to listen to it. A synopsis would be nice, along with a subject to discuss afterwards.

Blimps.

 

[Jeff Corey]

...You’re getting closer to the idea, but still missing some crucial elements. Can you answer this question: Why are basketball players like coins?...

Because neither show the "hot hand" effect. (Gilovich, Vallone & Tversky, 1985)

But what about golfers and dart players? http://homepage.psy.utexas.edu/homepage/group/GildenLAB/pubs/StreaksPsychonomicBR(1995).pdf

 

[Elizabeth I]

Why are basketball players like coins?

Why is a raven like a writing-desk?

 

[Sledge]

Why is a mouse when it spins?

 

[Rodney]

Are you saying that because we don't know about the non-coincidences we can discount them or that they really aren't happening?

I'm saying that there are countless coincidences -- as well as non-coincidences -- that we don't know about. For example, some time back, the Washington Post ran a story of a man who was sitting beside a stranger on the DC Metro (subway). The stranger noticed that the man was wearing a particular sweatshirt (I believe -- it may have been something else), which led to a conversation that resulted in the man discovering that the stranger was now living in the very house that the man lived in as a child. My point is that, generally speaking, the man would not have had a conversation with a stranger on the subway. In this case, because of what the man was wearing, a conversation took place, which led to the discovery of an amazing coincidence. So, how many times do strangers who may have something significant in common sit next to one another on the subway? We have no way of knowing.

 

[stevea]

The human brain evolved to look for significant relationships in order to make sense of the world. Coincidental events appear causal for that reason.

So you claim the brain (must have) evolved to seek out SIGNIFICANT relationships, therefore it is fooled by INSIGNIFICANT coincidental relationships. That statement makes little sense. It's a bit like saying we evolved to seek out nutritious food, therefore we also eat a lot of rocks. This also misses that the point of evolution is successful reproduction, not truth or knowledge except as a byproduct..

What we can infer is that either we haven't evolved sufficiently to have an innate sense of statistics or else that believing that coincidences are actually correlated (causal is too strong) does not extract a high reproductive cost. The latter seems unlikely - I mean throwing your best sheep onto a pyre when you lack rain or need some other random event to occurs seems like it would reduce fitness. Creating a "cargo culture" seems like an extravagant waste of energy without advantage. At least cathedrals and pyramids represented social systems that might confer advantage.

Humans seem innately blind to basic statistics - even basic averages are beyond many. Las Vegas and the State Lottery are proof of this.

Because neither show the "hot hand" effect. (Gilovich, Vallone & Tversky, 1985)

But what about golfers and dart players? http://homepage.psy.utexas.edu/homepage/group/GildenLAB/pubs/StreaksPsychonomicBR(1995).pdf

Coin tosses, performed well are nearly statistically independent events. The show presented argumentation but no evidence about basketball shots by a player being independent, one commenter claiming they were Independent, and another claiming they had negative correlation. Not only is the question easily resolvable by examining the data, but it deserves a hardy, "who cares ?".

We have perfectly adequate means of assessing statistical significance if you are willing to read a good undergrad stat book.. What is the mystery ?

 

[quarky]

This is vaguely on topic, I hope:

I used to build certain things that were suppossed to look "random", like uncut flagstone floors. It was very odd trying to come up with a precision sort of randomness. Three similar stones near each other, which would likely happen if the stones were dumped from the sky, simply looked too un-random. In creating non-patterns, there is some paradox.

 

[Skeptic Ginger]

So you claim the brain (must have) evolved to seek out SIGNIFICANT relationships, therefore it is fooled by INSIGNIFICANT coincidental relationships. That statement makes little sense. It's a bit like saying we evolved to seek out nutritious food, therefore we also eat a lot of rocks. This also misses that the point of evolution is successful reproduction, not truth or knowledge except as a byproduct.

What we can infer is that either we haven't evolved sufficiently to have an innate sense of statistics or else that believing that coincidences are actually correlated (causal is too strong) does not extract a high reproductive cost. The latter seems unlikely - I mean throwing your best sheep onto a pyre when you lack rain or need some other random event to occurs seems like it would reduce fitness. Creating a "cargo culture" seems like an extravagant waste of energy without advantage. At least cathedrals and pyramids represented social systems that might confer advantage.

You are starting with some false assumptions. You are assuming detecting a significant relationship is positive, equal to the negative result of assuming an erroneous causal relationship. It's possible correct deductions about causality are more beneficial outweighing the negative effect of getting it wrong.

For example, it would have been correct for ancient humans to correlate planting seasons with the movement of celestial bodies. That would have been a beneficial thing for a brain to conclude. Concluding the dead ancestors influenced those celestial movements because a drought occurred after the murder of a high priest may have been a benign conclusion.


In addition, you are not considering that the brain's tendency to draw conclusions about causality may result from some other beneficial brain function. Thus conclusions about causality could be the side effect of a different benefit in the brain's organization structure. The brain, for example, categorizes our environment. So when you see a tree species you've never encountered before, you know it is a tree because the brain naturally organizes certain data about trees, long before you encounter the new species. It is possible this brain organization incudes the effect of seeing causality because the brain is geared to see patterns.


To resolve these issues one needs first to describe how the brain organizes information and whether or not it does naturally seek to relate things by causality. Then one looks to explain how/why these brain traits were naturally selected. It's less efficient to start with the explanation and then fit brain function into the conclusions you've drawn about what should have evolved.

Humans seem innately blind to basic statistics - even basic averages are beyond many. Las Vegas and the State Lottery are proof of this....

And yet we've evolved brains capable of analyzing the inherent natural flaws in assessing evidence allowing us to use other tools, like statistics, to improve our causality conclusions.

 

[stevea]

But the other option is to assume the event was unique when you also don't know the number of failed incidences.

All events are unique. Many unique events have similarity. Sloppy language.

In creating non-patterns, there is some paradox.

Well there is certainly a paradox here. But I don't think your patrons really wanted "random" patterns as much an as aesthetically pleasing mixture of colors and textures. Perhaps that aesthetic says something about the human notion of random, but that's not in evidence.

You are starting with some false assumptions. You are assuming detecting a significant relationship is positive, equal to the negative result of assuming an erroneous causal relationship. It's possible correct deductions about causality are more beneficial outweighing the negative effect of getting it wrong.

You are having difficulty reading accurately. I explicitly said that the cost of an erroneous correlation may not be substantial, as one possible explanation. Clearly I did not assume that benefit/cost of correlation/coincidence were "equal" as you assert - I suggested just the opposite.

All creatures survive based on accurately correlating events. A grain of wheat refuses to germinate until it has seasoned several months and also has accumulated about 40% water weight. Those two events, the temperatures and perhaps others correlate well with the beginning of a reasonable growing season. Of course higher organisms can create dynamic/learned correlations.

There is nothing fundamentally different about any small set of data in favor of correlation vs evidence or only a coincidence. If you've never seen a coin, but you hear someone call out the results TTTHTTTTTHHTT, then you can't tell if the probability of heads is 25%, 50% or if probability of H is actually 75% and this is just a flukey-run. It's only when you examine larger sets of data that we can assign the extent of statistical significance.
http://en.wikipedia.org/wiki/Law_of_large_numbers

So here is the survival problem. We have an immediate need for a decision. Is there likely to be water at the stream 5 days from here - or should we stay put and wait for rain ? Are the buffalo likely to run in the next few weeks or do we want to head for the hills seeking antelope ? We are required to make a survival decision but the samples from the sample space may be too few and the number of dependent factors too large to give much statistical significance to one's "best guess". We all have the same dilemma today - tho' it comes in different forms.

Without a significant number of samples we cannot distinguish correlation from coincidence - that's normal, that's life ! But even when humans have a large set of evidence, enough to assess the statistical significance, we persist in supporting the theses created by coincidence. I don't want to divert the thread, but praying for rain, ghost dancing, snake-oil cures, homeopathy, and other forms of woo thinking are based on ignoring evidence of the statistical insignificance of the reported correlations..

Anyway humans seem to have a fairly crude "feel" for "averages" or expectation values, but we seem woefully oblivious to more detailed understanding of statistics.

Concluding the dead ancestors influenced those celestial movements because a drought occurred after the murder of a high priest may have been a benign conclusion.

Is this meant to be proof by hypothetical example ? What if, like the Maya you sacrifice thousands to appease the gods and lose your labor force ? What if like the Christian/Muslims you accept certain coincidences about some guy who lived millennia before and then go about killing thousands. What if you accept some exceedingly weakly supported claims of nationalist/genetic superiority - then go about killing others by the million ? What if you accept some dubious claims about the wonderful economic and social advantages of socialism, and the regimes end up killing tens of millions ? What if you believe it's your duty to bring about the second-coming and you have access to nuclear weapons ? Sorry - but the impact of incorrectly assessing correlation and coincidence isn't always so benign.

I'd argue that all religion and many social movements involve accepting poorly evidenced coincidental bits of evidence and extrapolating wildly.

Survival/reproduction is a species characteristic. Individuals making personally negative decisions may aid in the species survival. Such altruistic or cannon_fodder genes may have improved species survival.

In addition, you are not considering [...]

Why do you constantly try to tell others what they think in such a denigrating manner ? Have you any idea how annoying your personalized straw-manning (fallacy) is ? How much it contravenes any sensible discussion ?

that the brain's tendency to draw conclusions about causality may result from some other beneficial brain function. Thus conclusions about causality could be the side effect of a different benefit in the brain's organization structure. The brain, for example, categorizes our environment. So when you see a tree species you've never encountered before, you know it is a tree because the brain naturally organizes certain data about trees, long before you encounter the new species. It is possible this brain organization incudes the effect of seeing causality because the brain is geared to see patterns.

1/ Why do you say "causal" rather than "correlated" ?
2/ The brain is largely a pattern recognition machine and yes much of the PR mechanism has hard coded components (some language learning features for example). This isn't news.
3/ Of course we see patterns - but that doesn't mean we need to impute causality or correlation until sufficient evidence accrues. Eyesenk suggested that "learning too fast" (jumping to conclusions) was charateristic of neurosis, and there is more than a little to his theory. So far you have provided no evidence in support of your suggestion that imputing causality without sufficient evidence is a side-effect of some other advantageous process. It's clear that jumping to conclusions can be disastrous, so where is this other bigger advantage ?

It's less efficient to start with the explanation and then fit brain function into the conclusions you've drawn about what should have evolved.

YOU are the one who made the unsupported conclusion about how the brain actually evolved in post #2. Don't blame me for your error.

And yet we've evolved brains capable of analyzing the inherent natural flaws in assessing evidence allowing us to use other tools, like statistics, to improve our causality conclusions.

Only in the past ~300 years. It sure would have helped out a lot over the past 2 million years. Imagine all the virgins who wouldn't have been sacrificed and non-believers who would not have been slaughtered,the damaging woo-medicine that could have been avoided.

 

[Jeff Corey]

...We have perfectly adequate means of assessing statistical significance if you are willing to read a good undergrad stat book.. What is the mystery ?

I'm puzzled. This appears to be a response to my quoted post, but it doesn't follow.

 

[Rramjet]

Your OP citation is about randomness. That is not an answer to the question, "why" does a random event seem non-random.

So you can keep playing this game or you can actually carry on a discussion about the topic you've brought up.

BTW, the OP link goes to a page of clutter not many will bother to read. I had to Google the subject to get to a link to the actual podcast.

Thank you for that link. I am only "playing" here because I think that many people have misconceptions about the nature of randomness – what it actually looks like – quarky captured the essence of the “randomness” problem and how people feel about it neatly when he stated:

” I used to build certain things that were suppossed to look "random", like uncut flagstone floors. It was very odd trying to come up with a precision sort of randomness. Three similar stones near each other, which would likely happen if the stones were dumped from the sky, simply looked too un-random.”

…and he somehow felt there was a paradox involved.

When people come across events that seem not to be random (like seven tails in a row from a coin toss), they have this vague feeling like “something’s, not right here”… and because of that tend to want to “explain” the event in terms of something other that random chance. But what is actually occurring when they do that is a misconstrual of the true odds of the occurrence happening - given all the coin tosses that occur, it would be extraordinary if a run of seven tails did not occur somewhere at some time.

Confirmation bias adds to the problem because people positively seek out events that confirm their hypothesis (beliefs) and discount those that do not (and the more strongly held the belief is, the greater this psychological pressure to confirm is). For example in the “two Laura’s balloon” story it involved all the things that were similar between the two girls, but discounted (never mentioned) all the differences between them – which when added up – actually outweigh the similarities.

I’ll admit I could have introduced the topic more accessibly. There were many reasons for doing it the way I did. For example my own confirmation bias had a hand in it. My hypothesis was that many members posting in the JREF would simply not bother to listen to the podcast before leaping in with (probably erroneous) comments – possibly I was seeking confirmation of that. Also the JREF is supposed to be an educational foundation, and this podcast is most definitely educational - and thus completely worthy of being drawn to member’s attention – it is something that anyone who calls themselves a sceptic should listen to - and the best way to learn about something is to actually do the work yourself, rather than have it handed to you on a plate. If I had provided a “ready made” topic, people most definitely would have began commenting without listening to the podcast – and that is what I did not want to happen.

Anyway, thanks again for the link to the original show and thanks also to Jeff Corey for this addition:

Because neither show the "hot hand" effect. (Gilovich, Vallone & Tversky, 1985)

But what about golfers and dart players? http://homepage.psy.utexas.edu/homepage/group/GildenLAB/pubs/StreaksPsychonomicBR(1995).pdf

 

[Skeptical Greg]

Sledge:

Your ability to judge evidence is up there with Stevie Wonder's ability to judge paint colour.

Thanks for the spit all over my display ...

 

[Sledge]

Then this thread hasn't been a complete waste of time. My work here is done!
*throws smoke bomb to ground, disappears with swoosh of cape, pausing only to trip over cape on stairs*

 

[stevea]

I'm puzzled. This appears to be a response to my quoted post, but it doesn't follow.

Sorry for any confusion. WRT to your comment, my point is that independent vs dependent events and the resulting concept of stochiastic processes is a well established part of statistics. B.ball players may or may not demonstrate a similar independence as coins - it's not evidenced.

The last was a comment on the main topic. The measure of correlation and also statistical significance are a part of nearly every paper in experimental sciences. We could apply these to the "balloon story" but we'd rapidly realize that the similarity factors between the girls are a few among millions. It's hardly surprising that the ballon would be delivered to any nearby young girl with a name matching the one on the note. The rest is "data mining", looking for a dozen similarities from a list of millions of personal properties.

When people come across events that seem not to be random (like seven tails in a row from a coin toss), they have this vague feeling like “something’s, not right here”… and because of that tend to want to “explain” the event in terms of something other that random chance. But what is actually occurring when they do that is a misconstrual of the true odds of the occurrence happening - given all the coin tosses that occur, it would be extraordinary if a run of seven tails did not occur somewhere at some time.

Well said. But let me re-position the case a little.

When something odd happens, people start looking for correlated events, which may or may not be causal. I don't think having an explanation (the causal relationship) is always necessary, however it seems to satisfy some innate human emotional need - the appeal to curiosity. Younger ppl seem to have greater curiosity on average than adults, but may be a developmental/social issue. I think this is perhaps the driver to the development of various woo ideas. People want a satisfying explanation so badly that they make one up despite the lack of supporting evidence.

--

A personal example - I hunt mushrooms, and there are some types that I enjoy are fairly rare in my area, so it's a bit of a treat to find them. When I find one I note (almost at a subconscious level) the surrounding conditions. I've found for example that these are more likely to appear on a north facing slope, Some are more likely to appear around beech trees, certain types of soil and undergrowth are more likely to correlate and so on. There is likely a causal relationship to a few of these factors (slope, beech trees, soil type) but a non-causal relationship to others like the related undergrowth.

Without a good deal of scientific analysis it's impossible to show that these mushrooms are mycorrhizoreal, therefore dependent on beech trees in a causal manner, but they are only correlated with the various underbrush species.

Understanding the causal relationships (the explanations) vs the accurate correlations doesn't really help me in hunting mushrooms, tho' it would be critical if I was to try growing them or trying to create conditions for them to grow.

Now if I threw in a coincidental factor, for example that I find more of these on a Saturday afternoon - and if I acted on this coincidence by not bothering to look except on Saturday afternoons, then we'd have a false correlation that would both deprive me of potential gains, forms a special kind of woo and introduces a confirmation bias.

 

[Rramjet]

Sorry for any confusion. WRT to your comment, my point is that independent vs dependent events and the resulting concept of stochiastic processes is a well established part of statistics. B.ball players may or may not demonstrate a similar independence as coins - it's not evidenced.

I think the idea here was that if the long term average of basketballer’s abilities to “make” baskets is examined over many games, then their “hot streaks” conform to random variations around the mean of their “usual” performance. Thus in any particular game, the contention was therefore that a player on a “hot streak” should not necessarily be favoured over other players who might be available to shoot a basket (ie; that based on statistics, it is not rational to favour such a player).

I think the mistake made here is the drawing of conclusions about individual games from the long term average. Thus where I think you are correct stevea, is that such a statistical analysis does not account for in any particular game the fact that a player just might be experiencing a real enhanced ability. We all experience “off” days in whatever endeavour we participate in and at least part of the explanation can be attributed to physical things like tiredness, a niggling minor injury, etc, or psychological factors such as stress, depression, etc. On other occasions we are “on song” and all our faculties are working at peak performance. Thus in any particular game our “hot streak” might just BE genuine and thus other players will be justified in favouring us (over other players) to shoot the basket. Just that over the long term, we will naturally regress to the mean. Thus I think the program maker’s might have got it wrong here, possibly letting their own confirmation bias blind them to the reality. Each game is an independent event, yes, but individual performance is not.

 

[Jeff Corey]

But Gilovich, et al (1985) showed that individual players' outcomes in a game were independent events. For another analysis of the possible advantage of "feeding the hot hand", see this...
"In basketball, players believe that they should "feed the
hot hand," by giving the ball to a player more often if
that player has hit a number of shots in a row. However,
Gilovich, Vallone & Tversky (1985) analyzed basketball
players' successive shots and showed that they are
independent events. Thus the hot hand seems to be a
fallacy." http://www.hcrc.ed.ac.uk/cogsci2001/pdf-files/0152.pdf

 

[jakesteele]

Shermer addresses this as well. If billions of events occur every day, then 1 in a billion chance events are going to occur. Those 1 in a billion chance events will appear to be more than chance because we are unaware of the billion (-1) failures.

That doesn't disprove synchronicity, though. What it does is offer a much-more-likely-than-not explanation for it.

 

[SpuriusMaximus]

The problem with this reasoning is that we're largely unaware of both the coincidences and non-coincidences that are going on all around us. Dismissing a one in a billion coincidence on the grounds that there were undoubtedly a billion non-coincidences that we don't know about may be comforting to skeptics, but it is not evidence-based.

1. But ultimately it's irrelevant how many times it happened, or if any at all. Expecting that it either is somehow significant (e.g., loaded dice) or otherwise it must have also had a billion failures before, is the inverse gambler's fallacy. And there's a reason why it's a fallacy. A one in a billion chance is just as likely to happen on the first try as on the billionth try. It could even be attempted once and succeed, and then not even be attempted ever again, and it still not be anything else than sheer chance.

Basically even being unaware of the billion non-coincidences, and heck even their non-existence, still is not reason to assign some special significance to chance.

2. A lot of things are really not even coincidence in the first place, but we just assign some special meaning to the result. Feynman illustrated it the best with this quote: "You know, the most amazing thing happened to me tonight. I was coming here, on the way to the lecture, and I came in through the parking lot. And you won't believe what happened. I saw a car with the license plate ARW 357. Can you imagine? Of all the millions of license plates in the state, what was the chance I would see that particular one tonight? Amazing!"

A lot of what passes for amazing coincidences or significant signs or such, is really on par with that license plate.

E.g., let's say I roll 5 6-sided dice and the numbers come up 3, 1, 4, 1, 6. Wow, it's PI with 4 decimals, rounded to the nearest. What are the chances of that? It must be special somehow. Except it really isn't. It's a combination of numbers that I'm assigning some significance to, although for the dice it's really nothing special. And I could assign special meaning to anything else that comes up. E.g., 1, 2, 3, 4, 5, or maybe 2, 3, 4, 5, 6, or 2, 2, 3, 3, 3 (it's a full house in Yahtzee) or whatever else that I notice a pattern in. And if you basically allow permutations the chance of _something_ coming up that someone can assign a significance to, is actually pretty high and only limited by their own imagination.

Sometimes you can't show that billions of things without such significance also happened, simply because there aren't that many possibilities that can't be assigned some other arbitrary significance instead or aren't just as unlikely too.

3. A lot of things are higher probability than a lot of people would think.

E.g., out of the 22 people in a soccer game, chances are well over half that two would have the same birthday. But if you announced at a game that the goalkeeper of a team and one of the attack team of the other have the same birthday, it would sound like an unbelievable coincidence to a lot of people, and possibly even having some hidden meaning too. Just because they didn't do the maths.

E.g., the same kind of thinking got some people to believe that women somehow synchronize their periods to each other.

E.g., coincidences that boil down to "I dreamt a dove last night and I saw a dove today, so it must be some kind of premonition". By sheer virtue of the number of common things around you, and unless you only dream totally unusual things, and never retrofit your memory of dreams at that either, _something_ will exist in both the dream world and the real world. And to some people it even seemed significant enough to build a whole religion around it.

E.g., horoscopes, palm reading, bio-rhythm charts, and other techniques that just sound as an unbelievable coincidence, but are really generic enough to fit everyone around just as well.

You can't be aware of billions of cases where the horoscope didn't fit at all for each one that fit, or billions of soccer games without birthday clashes for every game with one, simply because there aren't as many in the first place. What looks like some unbelievable coincidence, in fact has a higher chance of happening than of not happening. The cases where it didn't happen, aren't billions for each case where it happened, but actually the minority set.