Here is our monthly place to discuss Overcoming Bias topics that have not appeared in recent posts.
Repeating a request for EY (or anyone else here, really), who listed Influence as a book so good and useful he keeps 2 lending copies.
Any other titles you do this with?
You really need to provide an RSS feed for comments. I believe WordPress makes this easy.
Right now, I am sitting in a class called “Literacy in Education” to complete my masters in education and I am noticing a fairly large bias. I’ll call it, treating means as ends. Kids are reading less and this is a major problem; kids should read more because reading is good. Presumably, if pressed, there would be rationalized justifications, but it seems like they’ve internalized the idea that reading is an end. Maybe there’s just a societal reading fetish in particular, but I think there is a more fundamental bias there.
Would be interesting if somebody reviewed _Predictably Irrational_.
Anybody know of any sites discussing problems and difficulties for independent learners, like in the comments on Initiation Ceremony?
Maybe something on the “incorrect feeling of understanding,” how to identify and avoid it.
Also maybe about the effects of logically unecessary or incorrect steps in a proof or argument that leads to a correct conclusion.
This is slightly off-topic, but people here have discussed being autodidacts a few times, and I was wondering if some of you would like to share your favorite textbooks, or what method you use to find good textbooks (maths, biology, physics, whatever).
josh, thanks for offering the only comment here so far that is not a request for someone else to do something.
I’m a lawyer. I’ve been tangling with “experts” peddling junk science for a number of years. They have pedigrees as long as your arm, can recite the literature backwards and claim to be adept Bayesians. They are able, using their expertise, to decide which studies have probative value and which do not. Then, relying only upon those studies that are valid and bear upon the issue at hand, they calculate a likelihood that the plaintiff’s malady was caused by the defendant’s product. Thereafter, they deduce that it was obvious all along that the product in question would harm the plaintiff decades later (though no literature existed to that effect at the time) – one even says he’s learned to overcome the hindsight bias and so is able to render opinions, to a high degree of certainty, about what long dead people thought would happen in the future. No kidding.
The only common thread I can find when reviewing their opinions is an absolutely perfect correlation between their causal and liability judgments and the contentions of the lawyer paying for their opinions. Weird, huh?
Anyway, I’d like to discuss how I might best deal with these sorts of “experts”. I’ve cited to trial judges everything from Pearl’s “Causality” to Cox’s “Algebra of Probable Inference” to papers by Gilovich and Kahneman and Rachlinski on the hindsight bias. All to no avail. “Let the jury sort it out; that’s what they’re here for” is invariably the answer. So, 12 people with an average of a 10th grade education get to decide what causes cancer and whether or not someone long dead knew it 50 years ago. Oh well, it’s a living.
Anything that might allow me to explain the niceties of all this to a judge with a political “science” degree would be greatly appreciated.
Well, as for my request, I’d be happy to do whatever work is required myself, if you want to give me access (and I meant Typepad, which I see is the underlying engine — recipe here). Real-world identity and credentials on request.
As for topics — I am interested in the social reputation problem as it pertains to intellectuals. In a world of millions of bloggers, along with credentialled academics, journalists, and writers, how do you choose who to believe on controversial questions, given that you don’t have time to investigate most questions deeply? Is there any way to streamline this process? One idea (not overcoming bias, but allowing biases to be made explicit and computed over) is here.
Would be interesting if someone posted about cats and how some of them are very cute.
I think an important way to overcome our biases is to give a fuller account of inference and causality. This blog seems to embrace the Bayesian perspective with an almost fundamentalist fervor.
I feel that some posts are victims of the Confirmation Bias, as they only look at the arguments to support the belief that the Bayesian approach is the ultimate solution.
What surprises me about this is that no one (posters or commenters) is challenging the Bayesian paradigm. This only reinforces the bias.
But maybe I got it wrong, and the use of the word *Bayesian* to denote someone’s intellectual position is supposed to mean “X believes that Bayes theorem is true” (which we all do, as it is a property of the conditional distributions).
Re: Bayesian paradigm. It’s hard to criticise subjective Bayesianism. But Solomonov Induction does seem to be highly overrated.
Eliezer has explained his Bayesian enlightenment of circa 2003.
I’d like him to expand on one aspect of this: His shift from seeing greater intelligence, and therefore the Singularity, as a good thing in and of itself, to extreme caution about Friendliness.
He has hinted about this, mentioning his realization that we should not enshrine things simply because they appear mysterious.
Still, I’d like to see more detail, both because of the theme of this blog and because of the centrality of the concept of Friendliness to his present research.
Thanatos: So, 12 people with an average of a 10th grade education get to decide what causes cancer and whether or not someone long dead knew it 50 years ago.
I’ve heard that some very experienced judges have said that, the longer they are on the bench, the more respect they gain for the decisions of juries.
A good Overcoming Bias topic: Is it true that juries, though untrained, really somehow overcome biases and come to relatively fairer decisions than other modes of deciding legal cases, and if so, how?
“My intention in this paper was to draw attention to the limitations of Bayesian decision theory by demonstrating its failure to cope even with the small move in the direction of realism that results from assuming that the decision-maker is not able to decide mathematically undecideable propositions.”
Here’s an interesting bias — every single one of OB’s contributors is male. How did that come to be? Is it a problem, and if so, how may it be rectified?
I can’t speak for the rest of the present Bayesians, but my use of the Bayesian approach is grounded in Cox’s theorems. The only serious attack on Cox’s theorems that I’ve ever seen was on a question of rigor, and was beyond my mathematical fortitude to follow. It turned on the exact assumptions we have to make about the belief function before the answer is uniquely Bayesian probability theory. Since I’m perfectly comfortable restricting myself to the class of differentiable functions (a much stronger restriction than necessary), I viewed the objection as a quibble and the eventual answer from mathematically inclined Bayesians as a counter-quibble.
Shorter me: when there are good arguments against my grounds for adopting the Bayesian approach, I pay attention.
As far as books I suggest/lend to others, I would have to say mine would be The Selfish Gene by Richard Dawkins. Dawkins seems to have an amazing ability to inject knowledge directly into the readers brain…
Michael — I’m only semiautodidactic, but a few things I’ve found helpful are finding out what people whose work I like choose to teach from or refer to, and digging into the works cited in other books read.
(As a side note, I use the Zotero Firefox extension to keep track of current and future reading material as well as research material; others may find it worthwhile also.)
Not only male, there is a heavy bias towards libertarianism (those two biases are not independent). Does this bother anyone, or are people happy with the default assumption that libertarianism is the path to objectivity?
The left has its own way of fighting bias (ie, Critical Legal Studies), which is very far removed from the style here. I wonder if there is any chance of fruitful discussion between these viewpoints?
Maybe slightly off topic, but funny nonetheless: They’re made out of meat
Actually I start to feel a little funny leaving my secnt marks comments around here without giving others any information on who t.f. is writing there, so I decided to write this introduction to myself. Skip it now, before it’s too late! =)
I’m a male (easy guess!) of 33 years (very close to 34) from Germany (I say this hoping that the geographic location where any person happens to be born is not as important as it used to be, and very much hoping that it won’t matter the least in glorious times to come!).
I’m educated as “technical assistant for physicists” (apprenticeship), “applied computer scientist” (german diploma, roughly equivalent to MSc or MPhil) and “master of science in information technology and automation systems” (MSc).
I am a self-employed, working as a technical consultant, mostly doing either simulation and control engineering, or just plain coding. My main hobby, since I shortly before I have started working, is machine learning (which will, shortly, lead to true, hard, smarter-than-though AI or I will eat a broomstick! [you may call me on this any time on or after the 1st of January 2040, if I should happend to live that long!]).
Beliefwise I am a bright (parentage is half protestant half old skool atheist).
Because of my rather diverse interests my bookshelf is aching under the weight of a plethorea of SF, F and technical books. From my “technical” shelf:
– anything by Richard Dawkins (get it! read it! essential and fun! [it gets a bit repetetive once you read nearly all of them])
– most of Daniel C. Dennett (surprisingly transparent and sane for a philosopher!)
– nearly all of Julien Offray de La Mettrie (a real hero to me! dull and obvious… now!, revolutionary and lethal [for him, so it seems] at his time!)
– some Willard van Orman Quine (admittedly, I haven’t yet read half of what I bougt)
– parts of Ludwig Wittgenstein’s “Tractatus Logicus Philosophicus” and “Philosophical Investigations” (I can be a rather fickle reader…)
– some Jerry A. Fodor (the completely wrong track, I think, but sure worth the time to check it out)
– the essential Douglas R. Hofstadter(“Goedel, Escher, Bach” and “Metamagical Themas” surely can’t fail to amuse)
– Richard S. Sutton & Andrew G. Barto “Reinforcement Learning” (the best introduction to RL you are likely to find)
– Julian Jaynes “The Origin of Consciousness in the Breakdown of the Bicameral Mind” (probably wrong, but surely a totally mindboggeling read!)
– Rober M. French “The Subtlety of Sameness” (quite subtle)
– Allen Newell “Unified Theories of Cognition” (quite gofai)
– Gerald M. Edelmann “Bright Air, Brilliant Fire” (to babbly for my tastes, but not without charme)
– some Fritjof Capra (god, I am open minded, ain’t I?)
– Jeff Hawkins “On Intelligence” (been there, read that)
– last but not least: …tloads of papers from Citeseer!!!
Btw I am not quite sober this evening (I visited my parents coz my it was my mum’s birthday yesterday), so please forgive me whatever you think needs forgiveness!
Peace and Rationality to all of us!
Regarding “Making Decisions in Large Worlds“:
I don’t claim to fully understand this material – so take my synopsis with multiple pinches of salt.
It seems to consider the case of observers with Godel sentences, and suggests that they will not behave as ideal Bayesians, incorrectly interpreting thier own inability to find evidence for their own Godel sentences as evidence that their Godel sentences are false – when in fact they ought to be reasoning that they are well out of their intellectual depth – and don’t properly understand what is going on.
This seems to be true, but no doubt some will say that the idea that finite observers are not ideal Bayesians is not news.
There’s also some “sour grapes” comments – about how Bayesianism pretends to be a solution to Hume’s problem of induction – and is therefore full of crap.
* Making Decisions in Large Worlds – Ken Binmore
* Can Knowledge Be Justified True Belief? – Ken Binmore
* An Impossibility Theorem on Beliefs in Games – Adam Brandenburger – H. Jerome Keisler
Cyan, note: Is Overcoming Bias Male? I have not sorted through the 100+ Google results for “libertarian site:overcomingbias.com”
At least for me, the division of the comments into multiple pages (when there are many comments) is quite inconvenient. Besides that it seems to prevent all the old links from working anymore.
Unknown, TypePad made that change without warning and without offering other options. We’re not happy with it.
Thanks for the pointer, Zubon.
HEY! I have a question. What is the painting used for the masthead on this website?
I don’t know if it’s too late to post on this thread, but . . .
One thing I have tried to get people interested in, is the apparently major influence of the choice-supportive bias in the current Democratic Presidential primary battle. Obama and HRC seem to have very similar political positions, and initially many, many Democrats didn’t have strong feeling either way; there were lots of people saying “I’ll vote for either one.” But, gradually, people made decisions (somewhat arbitrarily, and based on maybe one or two factors – anchoring?), and over the months that followed, people’s choices cemented. Now a huge percentage (double digits) of supporters of either candidate say they won’t vote for the other Dem. Their choice between two similar candidates has become so entrenched that they won’t even support the other Dem over a Republican, who is in reality MUCH more different from their beliefs than the alternate Dem is!
Also, bonus, I am female.
The feed at http://www.overcomingbias.com/atom.xml , which is in the feed-autodiscovery links in the HTML head, has silently stopped updating; the most recent entry in it is “Angry Atoms”. I only noticed this when I happened to realize I hadn’t seen any posts from this site recently; I expect many others will not have noticed.
At a minimum, it should be deleted (rather than remaining stale); better, it should be repaired.
See: John William Waterhouse: Ulysses and the Sirens – 1891.
are people happy with the default assumption that libertarianism is the path to objectivity?
It’s the other way around. The abandonment of certain biases is one path to libertarianism.
The same phenomenon seems to occur among Tarot card users: the longer they go on using the cards, the more they believe in the wisdom of the cards.
I’d like to request Eliezer do a series of posts intuitively explaining financial engineering investment algorithms, like the type used by Renaissance Technologies. It’s right up there as popularly inscrutable applied math with quantum mechanics and bayesian statistics.
In response to this post by MTravern:
“Not only male, there is a heavy bias towards libertarianism (those two biases are not independent). Does this bother anyone, or are people happy with the default assumption that libertarianism is the path to objectivity?
The left has its own way of fighting bias (ie, Critical Legal Studies), which is very far removed from the style here. I wonder if there is any chance of fruitful discussion between these viewpoints?”
Yes, the bias towards libertarianism here bothers me. In particular how Robin seeks framing health care/policy bias posts in the context of “hey, I’m kind of making this libertarian argument that less government funding/regulation of this or that would actually improve health or not harm it” on pretty much every post. At least it would be fun to see some self-critical meta-transparency by the regular OB contributors about how their biases warp their posts and their attempts to overcome bias. And critical legal studies (and its equivalents in the other social sciences, which heavily inform it) would be useful here. Particularly various analyses of economists, philosophers, physicists, and economists themselves, looking at how bias can skew scientific inquiry and the presentation of results.
Hopefully, I’m surprised you see my blog posts as obviously libertarian, especially on medicine. I have almost never taken positions here regarding more versus less government intervention. Do you, like James Hughes interpret my lack of explicitly endorsing more regulation as an endorsement of less regulation?
Robin, nope. I interpret you as doing a little dance where you seem honest about the data, but always seem careful to put some effort of framing the discussion of the data such that your posture is “here are ways that with less government regulation/funding we can maintain or improve on the health status quo”. Even though the data that you seemly to fairly present often seems equally framable as “here are some ways in which government regulation/funding DOES get better results than the lack of it”.
It seems, in my opinion to be dialectic-seeking, rather than enlightenment-seeking posturing. Sort of like, audiences seem more attracted to media that has a classic oppositional/dualistic undercurrent, such as individualist vs. collectivist, and often award more status and attention to participants in these dialectics, over people who put out media that seems more limited to empirical inquiry and problem-solving. I suspect this is a deeply rooted primate aesthetics. Perhaps an intuition that it’s more important of observing two alphas battle it out, and of choosing the right one’s side to be on, than it is to observe one’s natural environment and make decisions of how to survive in it without considering one’s relations with dominant and challenger alphas.
If I get a chance I’ll try to rewrite some of your recent and classic posts on this topic, as I imagine they could be without what I think is the nonexplicit libertarian/individualist/small govt. etc. posturing.
ps I ended up adding here and there to this post, without a lot of time to edit it overall, so my apologies for any disorganization or incoherence.
Apparently there is a completion bias, in which we’re more motivated to complete something close to finished, then to spend our energy in other areas that would benefit ourselves more. I’m concerned that completion bias may play a role in how scientists and thinkers are rewarded (are they rewarded more for finishing something, then sharing unfinished work? I don’t know of prestigious publication venues for partially finished work.)
Thus, Aleks Jekulon (sp?) in my opinion deserves praise for putting a lot of unfinished work of his on the internet for the rest of us to look at. Thoughts from the regular OvercomingBias contributors about completion bias and how scientists and thinkers are rewarded?
Aleks Jekulon’s website as a model:
Hopefully, I’m surprised you see my blog posts as obviously libertarian, especially on medicine. I have almost never taken positions here regarding more versus less government intervention.
I think even asking the question is considered to be ‘libertarian’; good liberals/conservatives don’t ask whether the policies they favor are effective, they’re supposed to know they are. Asking implies doubt and skepticism, which are sins.
Hmm … I seem to have posted this on an older open thread by accident. I’ll repost it here.
I’m wondering if Bayes’ theorem can be used to justify racial stereotyping. I’m hoping it can’t, for ideological reasons obviously, but someone I know recently made what seemed to be a reasonably strong case for it, which I’m finding difficult to refute because I don’t have a solid grounding in probability, but he does (or at least claims that he does).
If this is a misconception, please dispel it for me so that I can refer my friend to your explanation (that’s right: I’m asking the Bayesian gods to do it for me. Sorry, I’m just very busy right now, and I figure you’re up to the job, and this is a place where more people will see it).
Anonymous – was your friend attempting to morally justify, or just epistemically justify stereotyping? I don’t think Bayes has much to say about the former issue. See my old post on Racial Profiling (and especially Blar’s comment) for further discussion.
Anonymous, it makes me sad for ideological reasons too, but your acquaintance is correct: it is rational to use information about average group differences to make probabilistic inferences about an individual.
Of course, as you acquire information about an individual, the relevance of group information in the form of stereotypes quickly vanishes–and of course we need to worry about inaccurate stereotypes, and the possibility of stereotypes influencing behavior to the point of becoming “self-fulfilling prophecies”–but the basic point, that it’s not wrong to generalize if you do it right, holds.
I should add–anonymous, I would mention to your acquaintance that one of the many empirically demonstrated ways in which real people depart from Bayesianity is that “naive subjects do not distinguish between p(A|B) and p(B|A) in most circumstances” (Rational Choice in an Uncertain World). These can be very different quantities. So it might be the case that P(A|B) is high and P(B|A) is low, and a false stereotype forms that “As are generally Bs.”
Any overcomingbias headliners doping for cognitive advantage? There’ve been some articles on that recently in the media. Thoughts on those articles?
What does this mean? Anything rational? Risk vs. Gamble? Gamble here means taking a very large risk past a certain threshhold? Since this guy is a grad student in journalism, I think he should be held to a high standard in terms of lanugage use.
“In his closing arguments last week, the prosecutor, Charles Testagrossa, said, “We ask the police to risk their lives to protect ours.” I agree. But they shouldn’t have to gamble with them.
Kyle K. Murphy, a former lieutenant in the New York Police Department, is a graduate student in journalism at Columbia.”
Exponential growth is typically limited in some way, so that it becomes “S”-shaped, i.e., it approaches an asymptote. Such trends follow the differential equation: dN/dt = k•N•(1-L/N), so as N approaches the limit, L, the factor (1-L/N) approaches zero, thereby stopping the exponential component’s influence in the simpler equation dN/dt = k•N, whose solution was exponential. The “S”-shaped curve is pervasive in technological change, and to use simple exponentials is to argue that there are no limits to growth.
A good rule of thumb is: All trends have limits.
… be a charity angel.