Why did we evolve confirmation bias?

[athon]

At present in the small amount of spare time I have, I'm pulling together a heap of resources to try to write something on pseudoscience. So far I've addressed why our brains evolved the way they did, and I've suddenly come to the realisation that I have little more than conjecture on why we evolved confirmation bias. I can't find anything in the literature which might answer my question, either.

If we have an idea about how something works, it makes sense objectively to attempt to falsify it. Yet our brain does the opposite - it seeks only to find confirmatory evidence, in effect acting as an inductive machine.

Why did our brain evolve this bias?

Athon

 

[sol invictus]

I think it's because the consequences of failing to believe in a real correlation are often much worse than those of mistakenly believing in a false one.

If you watched a saber-toothed tiger eat your friend Uurg, it's much safer to believe that saber-toothed tigers are dangerous than that it was a fluke and they actually make good pets. If you get really sick after eating some strange mushrooms, much better to conclude they are dangerous than that it was a coincidence. The downside is religion and superstition, but those are less dangerous.

My $.02, anyway.

 

[jsfisher]

Why did our brain evolve this bias?

The ability to make generalizations is something of a survival skill. So, there is evolutionary pressure favoring developing the ability to move from the specific to the general. Isn't confirmation bias a continuation of generalization?

Consider this: A bad experience with one mushroom will tend to make me shy away from all mushrooms. Any tendency to falsify that generalization (the opposite of confirmation bias) can easily result it death.

Death, as a generalization, isn't a trait favored by evolution.


ETA: Sol invictus beat me to it.

 

[athon]

Yeah, I was tending in that direction.

I'm also thinking it has a social application - we seek out information that confirms those beliefs shared by a community.

Thanks for the thoughts.

Athon

 

[cj.23]

http://www.sscnet.ucla.edu/comm/haselton/webdocs/HaseltonFunder.pdf
may be of interest?
Also try
http://www.springerlink.com/content/ru83312771581275/fulltext.pdf

cj x

 

[athon]

Thanks cj. I have the Jones and Sugden one, which is damn useful. In fact that's where I've gotten most of my information from.

I'll look at the other one, though.

Thanks heaps!!

Athon

 

[learner]

I think it's because the consequences of failing to believe in a real correlation are often much worse than those of mistakenly believing in a false one.

If you watched a saber-toothed tiger eat your friend Uurg, it's much safer to believe that saber-toothed tigers are dangerous than that it was a fluke and they actually make good pets. If you get really sick after eating some strange mushrooms, much better to conclude they are dangerous than that it was a coincidence. The downside is religion and superstition, but those are less dangerous.

My $.02, anyway.

You took similar words out of my mouth. Good concise post.

 

[athon]

Second one is interesting. Unfortunately, it pretty much concludes by asking the same question I did.

In any case, the research seems to confirm much of my suspicion - confirmation bias (like most biases) result not from our brains not functioning properly, but rather functioning well for a social environment.

Sweet.

Athon

 

[articulett]

Yes... it gives us confidence to act. I think it's an outgrowth of associating correlation with causation (which is a nice generally useful algorithm--almost Pavolian, and useful for learning.) Once we form a hypothesis, (e.g."X causes Y"), we appear to seek evidence to confirm that belief and negate evidence that denies it.

It's fantastic when you are on the right track... you gather more and more useful information and hone it. But it can so easily go awry. For example, in many parts of the world, someone will start blaming disasters on "witches". They kill the accused witch and the disaster seems to go away. If a new disaster strikes, the next witch that needs killing is sought out. The witch killer gains status as a savior--a person protecting others from disaster, but those accused of being witches aren't so lucky. The savior gets to propagate his genes preferentially as a reward (and kills potential rivals in the process.)

I found this link on it: http://skepticsplay.blogspot.com/2008/09/confirmation-bias.html

 

[Jeff Corey]

Why did we evolve rapid learning? Contingent reinforcement leads on to positive outcomes and noncontingent renforcement leads to the establishment of superstitious behavior. Only looking for the card in the the Wason Card Selection Task that confirms the statement that "All cards with a odd number have an vowel on the other side" gets people to pick the wrong choices out of 1, 2, A, B. At TAM1, only 26 out of 156 got it right.

 

[Soapy Sam]

athon- are you sure that "we evolved confirmation bias" at all?

CB is either a mental state or a behaviour. There may be a genetic "setting" that biases us towards bias,in the extended phenotype sense, or it might be simply an emergent property of brains that happen to be built that way. The whole question seems a bit slippery.
To have a bias confirmed, first you need to have a bias in belief or expectation.
If bias in belief is simply expectation due to past experience, then it's hard to see how it can have "evolved". Once memory and pattern recognition exist, bias emerges automatically. Bias IS expectation. Foresight.

 

[athon]

athon- are you sure that "we evolved confirmation bias" at all?

CB is either a mental state or a behaviour. There may be a genetic "setting" that biases us towards bias,in the extended phenotype sense, or it might be simply an emergent property of brains that happen to be built that way. The whole question seems a bit slippery.
To have a bias confirmed, first you need to have a bias in belief or expectation.
If bias in belief is simply expectation due to past experience, then it's hard to see how it can have "evolved". Once memory and pattern recognition exist, bias emerges automatically. Bias IS expectation. Foresight.

I understand what you mean - I probably should phrase it as 'what led to the organising of our brain in such a way that allows us to experience cognitive bias?'

I'm more of a Dawkinsian than a Gouldist, so I've never really been convinced of spandrels of any type. However, even if CB is something of a cognitive spandrel, I'd be interested in how the underlying features which cause it developed, so it more or less amounts to the same thing.

You might have a point in that it is synonymous with expectation, yet this still doesn't address the question. Why, then, is this form of expectation of particular advantage that the cognitive tools which support it allow it to remain? We do have the ability to look for observations of both confirming and falsifying qualities - why does our brain only make a pattern with the confirming ones?

The papers cj linked to have led me through some interesting avenues, and they themselves are interesting in their own right. Although the research on the topic is pretty non-existent, most of the circumstantial evidence I can find on the matter does seem to indicate that most forms of bias are of some advantage in social systems (is it an irony that this supports my predictions? )

Athon

 

[Soapy Sam]

Next question then- what do you mean by "we"?
Humans?
Mammals?
Animals?

I'm pretty certain dogs expect things based on past evidence- and I've seen them baffled and obviously confused when their expectations were not met.

I would suspect that if there are neural structures involved , they antedate human brains by a very, very long time.

(Incidentally, I always preferred Dawkins to Gould myself, but this was at least in part because I found Gould's popular writing style over ornate and needlessly complicated. Dawkins assumes you know what a spade is. Gould gave you the history of the shovel in Regency England.)
But on the spandrel question I'm neutral. Some evolutionary features are convergent because there simply exists a limited number of ways nature can build something. The reason animals do not have wheels is not necessarily that it's impossible, but that it's impossible to get to wheels from where we are (and have been) in evolutionary design space.

The great complication for people seeking selective adaptationist explanations of behaviour is that it's hard to pin down the environment in which such an adaptation would be a best fit. It's the weakness of the meme theory.
Merely by existing, any adaptation changes its own environment . As the human environment is- critically- other humans, a change in behaviour may spread through that environment incredibly fast , changing it dramatically, with no genetic change whatever.

Incidentally- do you have data to support the view that our brain only makes a pattern with the confirming ones?

Is this not just reinforcement? If memory is created by reinforcing electrochemical pathways / potentials/ whatever in the brain, is it not to be expected that those which are confirmed will be reinforced while those that are proved wrong will be quietly dropped?

 

[paximperium]

How about Heuristics?

Heuristic (hyu-ˈris-tik) is a method to help solve a problem, commonly an informal method. It is particularly used to rapidly come to a solution that is reasonably close to the best possible answer, or 'optimal solution'. Heuristics are "rules of thumb", educated guesses, intuitive judgments or simply common sense.
In more precise terms, heuristics stand for strategies using readily accessible, though loosely applicable, information to control problem-solving in human beings and machines.[1]

http://en.wikipedia.org/wiki/Heuristics

 

[athon]

Next question then- what do you mean by "we"?
Humans?
Mammals?
Animals?

Good question. By 'we' I do mean people, however that is not necessarily to the exclusion of other species, or that I imply it occurred in a time between our split from apes and present day. Other animals could well experience something similar - it would be an interesting field of study, as if we could find some primary equivalence, it would certainly test my 'social' theory on the matter.

I would suspect that if there are neural structures involved , they antedate human brains by a very, very long time.

I'm fairly neutral on whether it would necessarily antedate humans. There would most probably be some domains which would play some sort of cognitive role in our far distant ancestry, which gave rise (directly or indirectly) to our biased way of making patterns. Whether it would be recognisable as the same neural setup we have now...

(Incidentally, I always preferred Dawkins to Gould myself, but this was at least in part because I found Gould's popular writing style over ornate and needlessly complicated. Dawkins assumes you know what a spade is. Gould gave you the history of the shovel in Regency England.)

Haha, I had a similar argument with a close friend of mine recently. She's a Gouldist, but then between you and me (I hope she doesn't visit here ) she could probably do well to have it explained to her what a spade is. With diagrams.

But on the spandrel question I'm neutral. Some evolutionary features are convergent because there simply exists a limited number of ways nature can build something. The reason animals do not have wheels is not necessarily that it's impossible, but that it's impossible to get to wheels from where we are (and have been) in evolutionary design space.

Oh, I fully agree. 'Why do we have X and not Y?' is a silly question sometimes - we don't have 'Y' because no pathway has been set up for it. Evolution is blind, as they say.

However, 'why X?' is, IMO, legitmate. Given variation and selection, there is always some sort of different way of doing things where one form deals with the environment better. Even today, we can see a range of cognitive abilities regarding bias, where some people experience it to a far greater extent than others. I'm not convinced it's always a learned behaviour, either, going on what I've read.

The great complication for people seeking selective adaptationist explanations of behaviour is that it's hard to pin down the environment in which such an adaptation would be a best fit. It's the weakness of the meme theory.
Merely by existing, any adaptation changes its own environment . As the human environment is- critically- other humans, a change in behaviour may spread through that environment incredibly fast , changing it dramatically, with no genetic change whatever.

I agree memetics does throw a spanner into the works, but as I suggested above, I'm not convinced that bias is purely cultural. That said, I'm also not suggesting by any means that it is simple. I'm inclined to see a feedback loop, where culture is influenced by a fundamental form of bias, encouraging that behaviour as a meme.

Athon

 

[athon]

How about Heuristics?

http://en.wikipedia.org/wiki/Heuristics

Yeah, the literature seems fairly keen on that as well, and I considered it as the most likely reason early on. The problem isn't so much that I'm short on theories - those I have plenty of. It's more that I was wondering if anybody knew of any research on the matter which might give me a little more grounding.

Still, it's good to simply get other's thoughts.

Athon

 

[Dancing David]

At present in the small amount of spare time I have, I'm pulling together a heap of resources to try to write something on pseudoscience. So far I've addressed why our brains evolved the way they did, and I've suddenly come to the realisation that I have little more than conjecture on why we evolved confirmation bias. I can't find anything in the literature which might answer my question, either.

If we have an idea about how something works, it makes sense objectively to attempt to falsify it. Yet our brain does the opposite - it seeks only to find confirmatory evidence, in effect acting as an inductive machine.

Why did our brain evolve this bias?

Athon

First off, lets us look to first principles before we get into the high level abstracted theories that have no basis to support them.

The brain basicaly sorts information and tries to make it match up, this is why confirmation bais exists, it inherent in the nature of the process. It is not something other thanw what the brain is supposed to do, it is a byproduct of the whole mechanism of association.

The brain looks for patterns, the brain creates patterns, so it just stands to reason that some patterns will have little validity. (And the brain manufactures perceptions wholesale)

Second, there is always going to be more information than the 'rational' side of the brain can process, focus and attention are important to human functioning, emotional flooding is a good example of what happens when there is too much information.
So with training and practice (like how ornithologists see birds everywhere , because they have trained where to look for them), we learn to sort out information. But we can also do so erroneously as well.

Third, a lot of processing happens at the preconscious level, it is emotional/intuitive in nature, mental patterns that just vaugely look like each other get matched up.

Fourth, operant conditioning, we have sets of signals, some are valid and some are not, The association of invalid and false signals is going to be reinforced at times.

Sixth social and cultural mores and precepts are learned from an early age.

 

[GreyICE]

Putting together clues is an immensely useful survival trait. As an anecdotal thing, imagine realizing there's no birds, seeing a slight movement in the foliage, and smelling something odd, and putting together the fact you're about to get jumped by a predator. Better safe than sorry is a survival trait.

Less anecdotally, consider that even dogs and mice have strong pattern-recognition engines in their brain. Pavlov, the infamous mazes, etc.

 

[athon]

The brain basicaly sorts information and tries to make it match up, this is why confirmation bais exists, it inherent in the nature of the process. It is not something other thanw what the brain is supposed to do, it is a byproduct of the whole mechanism of association.

The brain looks for patterns, the brain creates patterns, so it just stands to reason that some patterns will have little validity. (And the brain manufactures perceptions wholesale)

I'm still yet to buy the 'inherent in pattern making' hypothesis. I'm happy enough to see it as a heuristic or risk-limiting strategy, where it saves energy to automatically bias this selection, yet the first-principles pattern-matching does not automatically assume a bias for the established model. I realise our brain follows a Bayesian form of problem solving, which demonstrates a bias to the majority, yet this is not the same thing as a cognitive bias for an already accepted model. Research demonstrates that we will actively seek out confirming information, not simply rely on Bayesian probability.

Second, there is always going to be more information than the 'rational' side of the brain can process, focus and attention are important to human functioning, emotional flooding is a good example of what happens when there is too much information.
So with training and practice (like how ornithologists see birds everywhere , because they have trained where to look for them), we learn to sort out information. But we can also do so erroneously as well.

Which is what others (including myself) have said above - it's saving energy. I think it's likely that playing the odds this way reduces risks and saves energy, while opening the way for some potential errors. All well and good. However, I was wanting to know if this conjecture had any supporting research.

Third, a lot of processing happens at the preconscious level, it is emotional/intuitive in nature, mental patterns that just vaugely look like each other get matched up.

True, but that's not really saying much.

Fourth, operant conditioning, we have sets of signals, some are valid and some are not, The association of invalid and false signals is going to be reinforced at times.

Sixth social and cultural mores and precepts are learned from an early age.

Thanks for the opinion, but I'm a little lost as to where we went in the end. I've got my own pet ideas, which are based on some pretty solid ground, but I was asking if anybody had anything solid on what factors might have influenced the evolution of this way of thinking over, say, something purely Bayesian or being able to falsify intuitively.

Thanks to those who gave some links, and to Soapy for the helpful criticism on my starting premise. It's much appreciated.

Athon

 

[Dancing David]

I'm still yet to buy the 'inherent in pattern making' hypothesis. I'm happy enough to see it as a heuristic or risk-limiting strategy, where it saves energy to automatically bias this selection, yet the first-principles pattern-matching does not automatically assume a bias for the established model. I realise our brain follows a Bayesian form of problem solving, which demonstrates a bias to the majority, yet this is not the same thing as a cognitive bias for an already accepted model. Research demonstrates that we will actively seek out confirming information, not simply rely on Bayesian probability.

When you consider that perceptions are manufactured it is important. Now some consider it trivial, I don't.

If the fovea is very small compared to the visual field, where do all the colors come from? The blind spot is relatively huge in the visual field, it gets filled in.

I consider that significant, our brains are used to filling in missing spaces. So I think that confirmation bias would fall into that.

And the superstition behaviors in operant conditioning would account for a lot.

I don't think that natural selection would select against confirmation bias until it started to increase the death of the organism.

Which is what others (including myself) have said above - it's saving energy. I think it's likely that playing the odds this way reduces risks and saves energy, while opening the way for some potential errors. All well and good. However, I was wanting to know if this conjecture had any supporting research.

And again, if you look at attention, there is just way to much information to pay attention to it all. So we sort through stuff, fortunately mostly at the preconscious level

But I don't see the confirmation bias as playing a role that rises to the level of effecting reproductive success. So, if there is some signal distortion, it does not effect the outcome.

I don't think it was selected for or against.

True, but that's not really saying much.

Yes, but if due to spurious association it gets reinforced, then you might find that the organism believes in silly things. without there being a specific mechanism to the effect.

The phone always rings when I am on the toilet could just be, I have a higher level of arousal when the phone rings and I am on the toilet (with a land line) because I am frustrated that I can't answer it. If the system is in a level of arousal, we may be more likely to recall that association.

Thanks for the opinion, but I'm a little lost as to where we went in the end. I've got my own pet ideas, which are based on some pretty solid ground, but I was asking if anybody had anything solid on what factors might have influenced the evolution of this way of thinking over, say, something purely Bayesian or being able to falsify intuitively.

Thanks to those who gave some links, and to Soapy for the helpful criticism on my starting premise. It's much appreciated.

Athon

The question is more to my mind, why would natural selection select for or against confirmation bias? I see it as a secondary effect.