Thinking fast, slow, and not at all: System 3 jumps the shark | Hacker News Reader

Thinking fast, slow, and not at all: System 3 jumps the shark

statmodeling.stat.columbia.edu

64 pointsmartingoodson13 comments

This is just embarrassing. The claim in the book about an absence of correlation implying an absence of causation is correct. The car thing is not a counterexample, simply because pushing on the gas pedal does not cause the speed to increase, obviously. For example, the car might be switched off. There is confusion about the concept of cause here.

Further embarrassment arises when the author talks about the lead book author’s credentials. Does he not even know who Kahneman is? It’s bad and irrelevant enough to dwell on what you think are an author’s qualifications, but at least find out something about him first.

You mean the same Kahneman who got:

- A Nobel Prize in Economics. - Whose book has almost 20,000 reviews with an average of almost 5 stars. - Whose book was winner of the National Academy of Sciences Best Book Award in 2012

and then ...

got most of the findings undermined by this blog analysis:

"Reconstruction of a Train Wreck: How Priming Research Went off the Rails" https://replicationindex.com/2017/02/02/reconstruction-of-a-...

and then...

commented as himself in the same blog https://replicationindex.com/2017/02/02/reconstruction-of-a-...

stating and I quote : "I accept the basic conclusions of this blog..."

You mean that one ?

In case you are thinking anybody could create an account under his name ...I provide you with an additional reference: https://retractionwatch.com/2017/02/20/placed-much-faith-und...

Lets just accept most of Academy nowadays is not far from a total joke.

He's a psychologist who wrote his book prior to the replication crisis in social psychology. I read "Thinking fast and slow" with a companion piece that detailed the most flawed parts of the book. I read that blog post before I read the book. There's basically no psychology you can read without some degree of BS in them so I love reading books which have already been thoroughly scrutinized. Some chapters were very well grounded, others like ch4 I didn't even bother reading.

I don't know where you got the impression that MOST of his research got undermined by that blog post when that blog post mostly took down a few chapters. If anything Kahneman impressed me for coming out of the replication crisis with not much more than a blown off limb. It would be great if Kahneman got everything perfect the first go of things but it was practically the point of his book that it's incredibly difficult to not make the exact mistakes that he ended up making.

The above posters general point is correct that the author doesn't really understand Kahneman's work, otherwise they would not have phrased things the way they did.

Just wanted to let you know that although I do not entirely agree with them, I appreciated both yours, and leephillips ( just above), well thought and considered responses.

Mind sharing that companion piece, please?

I am also trying find that companion piece. Maybe this[1] is what was being referred?

[1] https://replicationindex.com/2020/12/30/a-meta-scientific-pe...

Yes basically that and the earlier linked article is what I had open.

Yes, that one. You exaggerate when you say that “most” of the book’s findings were undermined, but a good chunk of them were, including most of the principal findings of one of the chapters. And note that Kahneman had the integrity and objectivity to accept that he had been taken in by some non-replicable results, and has been totally clear, honest, and straightforward about what results that he described in his book are reliable and which should probably be discounted. Maybe he was too credulous when he wrote the book, but no one realized at the time how much of social psychology would collapse. He has provided a model of how an academic should react when faced with uncomfortable evidence.

His response to that blog is really impressive. It must be so hard to look at significant bodies of work that you've done and have to reevaluate it. Even more so when you're probably the most famous living figure in social psychology.

I would hope that in the same circumstances I would do the same, but I can't honestly say that I would live up to that standard.

One of the examples given on Twitter that I found enlightening (not an expert so I hope I'm interpreting this correctly): if you have a homeostatic system, you would specifically expect to find less correlation between the components than between causally unrelated quantities (e.g. when the correlation between the position of a chicken's body the position of the same chicken's head is lower than the correlation between the position of the chicken's body and a different chicken's head: https://imgur.com/gallery/vgcxL)

From the post:

> I’m sure that my above post is unfair in the sense that these three people spent several years working hard on a book, and I’m basing my entire reaction on some combination of the title, a technical error that someone found, and an interview where one of the authors was maybe a bit too relaxed. These three pieces of information are in no way a summary of the actual book!

And here we see System 1 in action, helping write a reaction to a book not even read yet.

I would argue that in the driving example, there is in fact correlation, but the data set (a single drive) is insufficient to detect it. If you analyzed the data from a thousand different drives over the same road, the correlation between speed and gas pedal pressure would indeed become apparent. Perhaps the book's authors might better have said that if causation exists, there must exist a set of measurements that can reveal the corresponding correlation.

Or are there cases where causation exists, but no amount of data will reveal correlation?

I was inspired to go into psychology after reading Malcom Gladwell, Freakonomics, et al. Net-net the degree was disapointing. In some ways an intuitive sense of people is superior to what the field's been able to discover. 'Noise' is a good descriptor.

Thank god I transferred into CS.

Reply on news.ycombinator.com