Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

You Should Downvote Contrarian Anecdotes

thobbs.github.com

Post reply on HN
Print view View original

71–80 of 158 posts

Re: You Should Downvote Contrarian Anecdotes

#71

Post by DanielBMarkham »

I had to vote this up because of the meta-nature of the article. If anybody makes a comment disagreeing with it while sharing an anecdote they only play into's the author's theme! It's like accusing somebody of being in denial: there's nothing they can say in return that doesn't somehow support the idea that they are in denial. Love it.

Having said that, this is another really bad article in what seems like an endless series of bad articles on HN. Here are a few of the more obvious flaws:

- There is no universal truth as the author seems to imply. Simply because some publication or source you may like has performed some sort of statistical study doesn't have a lot of meaning on it's own. Yes, 100% of the people who eat bananas are dead within 120 years of their consumption. No, that does not mean bananas are bad. The study or scientific reporting is simply the beginning of a much longer conversation society has over many decades that leads us to higher-fidelity models.

- The purpose of a social site is to behave socially. While places like HN have (or used to have) a lot of different guidelines for the types of behaviors that are encouraged or not, being social means sharing stories, anecdotes. We are not robots.

- The idea that people are unable to sort out personal anecdotes from other forms of information. The follow-up idea that since they are not able to do this, we should prevent ourselves from sharing such stories. This is bad, bad, bad, bad, bad, bad .... bad(n). We are humans. We share stories. Anybody who says "people are so stupid" can justify just about damn near anything as long as they keep emphasizing the stupidity and danger some people's actions represent.

Every now and then, gasp(!), published research is either wrong or doesn't show anything near what the reporter claims. Anecdotes don't help with this, of course, but they serve to remind us that even well-known scientists working at the highest quality standards available are still just sharing with us a very specialized form on anecdote. We did these things in this way and this is what we observed. Here is how you also can observe this. The really "good" part of the story they are sharing is talking about hidden assumptions, population variance, reproducibility, and so forth. Anecdotes don't do this, but they help us brainstorm ways in which we can improve the discussion, take the next experiment to an even better place.

I'm very uncomfortable with the line of reasoning that goes somewhat like this: people are broken in some way, therefore we must somehow control what they read, say, or think for their own good. To me the beauty of western civilization is that really broken people can do these amazing and awesome things. The fact that we're deeply flawed is the magic. Science and human advancement work because of our flaws, not in spite of them. This is a very important thing to understand! Setting up some ideal of perfection, no matter how well-intended, and then mucking around with the way societal interaction works in some effort to improve on things is heading down a very dark path that has a very unhappy ending. This attitude seems rife in the technology community, however, perhaps because we are such analytical people.

I don't want my fellow man to be irrational and distrustful of science and knowledge. But I'll take that any day over silencing contrarian articles and dissent. We've done the math on this: wrong people who share emotional stories and persuade crowds about all sorts of illogical things are a price that a dynamic community pays for progress.

Re: You Should Downvote Contrarian Anecdotes

#72

Post by refurb »

polyfractal wrote: post #66
retrogradeorbit wrote: post #28

In a world were 47 of 53 landmark cancer research papers were found to not be replicable... where one scientist admitted (after the meta-researchers tried 50 times to replicate the result and failed) that they had run the experiment 6 times and got the result once and reported that because it made a better "story"... where the act of suppressing refutations of other's bogus research is "culturally commonplace"... Som…

This is utter hogwash. I'm as critical as anyone (probably more so, check my comment history) of academic biology because of my background in it. There are certainly things wrong it. And due to the nature of biology, replicating results is really hard . It's a fact of life when you deal with systems that are not perfect, not identical and very opaque. But to say that "Science has been supplanted by money and politics…

While I agree that "Science has been supplanted by money and politics" is going way overboard. I'm going to take the other side of the argument because I think you are way off base. I worked in both an academic lab and a "big pharma" lab(4 years and 6 years respectively).

To say "most big pharma labs" do not have access to the literature is laughable. We had better access than most academic institutions. If we needed a paper we didn't have access to, it took a few hours to get it. The company was more than willing to pay the $50 to get a copy of whatever paper, since we would often blow $50 running one experiment. Many of the smaller biotech might have poor access to journals, but even then, if you could justify the cost, you could get it.

Second of all, yes I trust labs that are trying to recreate data to make a drug out of it. You have to remember that these attempts to recreate data were a very important data point on a potential multi-million (billion?) dollar investment in a new target, these are NOT low priority projects. They WANT the data to be true. They have zero incentive for the data to not be reproducible.

Having worked in both academic and commercial labs, I would say the incentive to "tweak" results in much great in academic labs for the following reasons:

1) Often results are never double checked in an academic lab unless the work is use in a later project. Contrast this with a pharma lab where if the data is positive, you'll have to prove it again and again. 2) Academics (both profs and students) live and die by papers, not so in academic (in fact, in the company I worked in, they preferred if you didn't publish) 3) Work in academic is often performed by relatively inexperienced ungrad and grad students, while big pharma scientists often have years of experience.

Re: You Should Downvote Contrarian Anecdotes

#73

Post by carbocation »

JoeAltmaier wrote: post #55
jessedhillon wrote: post #22

Earlier quoted context omitted.

In your example, A is not relating a statistical or experimental finding, but instead dispensing commonsense knowledge which happens to be objectively false. It's not what the OP is talking about.

Commonsense, like ulcers are caused by diet and anxiety, or that salt raises blood pressure? I think a good point is made: statistical evidence is also misleading - it deliberately ignores (averages out) the extreme cases. The results are a distribution; statistics folds that into one number. Anecdotes fill out the distribution.

Ulcers are not generally caused by diet and anxiety, but rather by bacterial infection. Anecdote is a much more compelling refutation of "common sense" than it is of empirical statistical evidence.

Statistical evidence is not misleading; it's simply the case that if you are seeking outliers, as in your example, then looking at measures of central tendency won't contain what you're looking for. Anecdote has no role in "fill[ing] out the distribution."

Re: You Should Downvote Contrarian Anecdotes

#74

Post by hexagonal » 

  For example, if a close friend goes on and on about how the Ford he bought 
  was a piece of crap, detailing how the transmission failed at 30k miles 
  and the rear-view mirror fell off, you’ll be wary about buying a Ford in 
  the future, even if Consumer Reports rates them highly.
Personal anecdote: I've got a '93 Ford Explorer. It has 237,000 miles on it, and has yet to have a single major component failure. I've been waiting for the goddamn thing to die for the last 70,000 miles, so I can buy a car that does better than 13 miles to the gallon, but no dice.

Don't buy Ford trucks. They're too reliable.

Re: You Should Downvote Contrarian Anecdotes

#75

Post by Spooky23 »

I guess it depends on the story. If HN is supposed to be about startup culture and technology, the anecdote is essential to being vibrant. What is a startup if not a story whose premise is unproven?

Now if you're telling me that your sister stuck a magnet in her ear and cured her cancer, I'm not going to give that credence without some real data.

Most things on HN are not that, and a good story often compels us to think. So FWIW, I'm not downvoting contrarians, and I'm not downvoting anecdotes for being anecdotes.

Re: You Should Downvote Contrarian Anecdotes

#78

Post by carbocation »

JoeAltmaier wrote: post #68
pgsandstrom wrote: post #57

Earlier quoted context omitted.

What do you think the voting system on HN does? It promotes good content, thus also demotes bad content. The author is saying "Here is a reason that content X is bad. If you agree with me, downvote it".

Doesn't make sense. If the content is 'bad', moderator could expunge it. Or writer could be counseled on giving better posts. Or any number of things. The voting system on HN sometimes demotes bad content. Far more often (discussed widely elsewhere) its used to assert agreement or disagreement. When used in this sense, its a sort of instant-poll. In the case of anecdotal argument, it forms an ad-hoc 'scientific' expe…

> Its statistical significance may exceed a graduate student's narrow study.

And thanks to selection bias, its external validity is likely zero. "Our HN poll found that 1 in every 10 people has founded and sold a company."

Re: You Should Downvote Contrarian Anecdotes

#79

Post by JackFr »

retrogradeorbit wrote: post #24
loup-vaillant wrote: post #23

Earlier quoted context omitted.

Let's take an example, since they seem to work best: vaccines. Vaccines protect you from the risk of contracting particular diseases, some of which are crippling, lethal, or incurable. Plus, most are extremely effective: once you take your shots, you are effectively immune. That's good. There is a downside however. Sometimes, vaccines have side effects. Most side effects are quite benign, but if you're unlucky enough…

Are you actually arguing against giving anecdotes weight by using... anecdotes?

Those aren't anecdotes, they are hypotheticals. They are useful, but statistically they occur even more infrequently than anecdotes.

Re: You Should Downvote Contrarian Anecdotes

#80

Post by alan_cx »

Please forgive me if I have fundamentally misunderstood this, but is this article suggesting that anecdotes are some how comparable to scientific research data?

If one wants to "attack" an anecdote, then a contrarian anecdote is the weapon.

If one wants to attack scientific data, you need contrarian scientific data.

At least, I hope that is right. There for, to mix the two is like attacking a tank with a wooden stick.

Surely, anecdotes are used as the premise of scientific research. Lots of people tell stories. There seems to be something interesting going on. Then you do the research and produce the data. If the data is conclusive, then your past the anecdote. If later contrarian anecdotes appear, and they seem significant, off you go to scientific research again.

I know Im wrong somewhere. But where?

Post reply on HN