Dean Eckles
@eckles.bsky.social
networks, contagion, causality
faculty at MIT
faculty at MIT
boxd.it/beN8Br
Thematically related...
Thematically related...
October 4, 2025 at 4:59 PM
boxd.it/beN8Br
Thematically related...
Thematically related...
Here's a sign at a ferry terminal on Martha's Vineyard. When do you think this was posted?
Presumably after 9/11, but with the date to make clear the cause.
Presumably after 9/11, but with the date to make clear the cause.
August 25, 2025 at 7:56 PM
Here's a sign at a ferry terminal on Martha's Vineyard. When do you think this was posted?
Presumably after 9/11, but with the date to make clear the cause.
Presumably after 9/11, but with the date to make clear the cause.
I wonder what personality psychologists think of this analysis on.ft.com/45G0aEF by @jburnmurdoch.ft.com (which updates this paper from a couple years ago doi.org/10.1371/jour...)
August 8, 2025 at 7:42 PM
I wonder what personality psychologists think of this analysis on.ft.com/45G0aEF by @jburnmurdoch.ft.com (which updates this paper from a couple years ago doi.org/10.1371/jour...)
Maybe American Time Use Survey could be useful.
The Pew question has been asked before, but I wonder if this sample is too small to have enough precision for this.
The Pew question has been asked before, but I wonder if this sample is too small to have enough precision for this.
August 4, 2025 at 7:42 PM
Maybe American Time Use Survey could be useful.
The Pew question has been asked before, but I wonder if this sample is too small to have enough precision for this.
The Pew question has been asked before, but I wonder if this sample is too small to have enough precision for this.
There are class differences in reading books in the US.
www.pewresearch.org/short-reads/...
I'd be fascinated to know whether these differences are getting larger or smaller. Are there good sources here?
www.pewresearch.org/short-reads/...
I'd be fascinated to know whether these differences are getting larger or smaller. Are there good sources here?
August 4, 2025 at 7:39 PM
There are class differences in reading books in the US.
www.pewresearch.org/short-reads/...
I'd be fascinated to know whether these differences are getting larger or smaller. Are there good sources here?
www.pewresearch.org/short-reads/...
I'd be fascinated to know whether these differences are getting larger or smaller. Are there good sources here?
Nice new paper in PNAS providing further evidence that random/long ties help social contagions — even many that would be labeled "complex contagions"
www.pnas.org/doi/10.1073/... @davidlazer.bsky.social
www.pnas.org/doi/10.1073/... @davidlazer.bsky.social
July 16, 2025 at 12:06 AM
Nice new paper in PNAS providing further evidence that random/long ties help social contagions — even many that would be labeled "complex contagions"
www.pnas.org/doi/10.1073/... @davidlazer.bsky.social
www.pnas.org/doi/10.1073/... @davidlazer.bsky.social
Threads cut off that second, key word here, so I started thinking, I guess people really hated Clippy
July 9, 2025 at 12:56 AM
Threads cut off that second, key word here, so I started thinking, I guess people really hated Clippy
Is this large of a difference also more common when there is a big change in government employment?
July 3, 2025 at 5:30 PM
Is this large of a difference also more common when there is a big change in government employment?
Say it was 1/4th the size — can this data reject that? Seems like maybe no? (squints at Figure 4)
Of course I get that probably prior arguments have made it seem like these effects would be giant, but presumably they've got be be importantly smaller than the focal issue effects.
Of course I get that probably prior arguments have made it seem like these effects would be giant, but presumably they've got be be importantly smaller than the focal issue effects.
July 1, 2025 at 2:08 PM
Say it was 1/4th the size — can this data reject that? Seems like maybe no? (squints at Figure 4)
Of course I get that probably prior arguments have made it seem like these effects would be giant, but presumably they've got be be importantly smaller than the focal issue effects.
Of course I get that probably prior arguments have made it seem like these effects would be giant, but presumably they've got be be importantly smaller than the focal issue effects.
I went over to X to check this (and clicking through a couple "This Post is from an account you muted" banners). I guess the poster here edited these screenshots, perhaps explaining the unequal gaps to to the timestamp line.
June 11, 2025 at 12:08 AM
I went over to X to check this (and clicking through a couple "This Post is from an account you muted" banners). I guess the poster here edited these screenshots, perhaps explaining the unequal gaps to to the timestamp line.
Sometimes nice slices of sharing on the internet appear in forums for specific (e.g.) audio equipment
June 10, 2025 at 12:04 AM
Sometimes nice slices of sharing on the internet appear in forums for specific (e.g.) audio equipment
Made me think about various things downstream of BLS's Consumer Price Index.
Two very direct examples: Social Security cost-of-living adjustments and inflation-protected bonds (TIPS)
www.wsj.com/economy/cpi-...
Two very direct examples: Social Security cost-of-living adjustments and inflation-protected bonds (TIPS)
www.wsj.com/economy/cpi-...
June 5, 2025 at 3:17 AM
Made me think about various things downstream of BLS's Consumer Price Index.
Two very direct examples: Social Security cost-of-living adjustments and inflation-protected bonds (TIPS)
www.wsj.com/economy/cpi-...
Two very direct examples: Social Security cost-of-living adjustments and inflation-protected bonds (TIPS)
www.wsj.com/economy/cpi-...
Obviously there's plenty to debate here, but so many responses to this are at the level of industry propaganda: "IP" is entirely the same as personal property
May 27, 2025 at 3:18 AM
Obviously there's plenty to debate here, but so many responses to this are at the level of industry propaganda: "IP" is entirely the same as personal property
If you want to do cost-benefit calculations (as this example paper does) you'll need more than a stat sig result: you need a more precise estimate.
Many stat sig results correspond to largely uninformative 95% CIs about relative magnitude
statmodeling.stat.columbia.edu/2024/03/07/r...
Many stat sig results correspond to largely uninformative 95% CIs about relative magnitude
statmodeling.stat.columbia.edu/2024/03/07/r...
May 3, 2025 at 4:10 PM
If you want to do cost-benefit calculations (as this example paper does) you'll need more than a stat sig result: you need a more precise estimate.
Many stat sig results correspond to largely uninformative 95% CIs about relative magnitude
statmodeling.stat.columbia.edu/2024/03/07/r...
Many stat sig results correspond to largely uninformative 95% CIs about relative magnitude
statmodeling.stat.columbia.edu/2024/03/07/r...
It does seem like some unexpected challenges came up in executing this RCT, but also that it was underpowered for more plausible effects on absences even if everything went as planned.
dagliano.unimi.it/wp-content/u...
www.socialscienceregistry.org/trials/11960
dagliano.unimi.it/wp-content/u...
www.socialscienceregistry.org/trials/11960
May 3, 2025 at 4:07 PM
It does seem like some unexpected challenges came up in executing this RCT, but also that it was underpowered for more plausible effects on absences even if everything went as planned.
dagliano.unimi.it/wp-content/u...
www.socialscienceregistry.org/trials/11960
dagliano.unimi.it/wp-content/u...
www.socialscienceregistry.org/trials/11960
Wow! Reducing absences by over 10%!
Oh but seems like the 95% confidence interval includes increasing absences! And effects over double this size.
(And this analysis deviates from the planned analysis in the pre-registration.)
Oh but seems like the 95% confidence interval includes increasing absences! And effects over double this size.
(And this analysis deviates from the planned analysis in the pre-registration.)
May 3, 2025 at 4:04 PM
Wow! Reducing absences by over 10%!
Oh but seems like the 95% confidence interval includes increasing absences! And effects over double this size.
(And this analysis deviates from the planned analysis in the pre-registration.)
Oh but seems like the 95% confidence interval includes increasing absences! And effects over double this size.
(And this analysis deviates from the planned analysis in the pre-registration.)
Interesting example of different intuitions about effect sizes.
@davidshor.bsky.social, who has run a lot of randomized experiments:
"You say the 70 words to him and then ... there's a 2.5% change he changed his mind. I think that's incredible!"
www.youtube.com/watch?v=Sx0J...
@davidshor.bsky.social, who has run a lot of randomized experiments:
"You say the 70 words to him and then ... there's a 2.5% change he changed his mind. I think that's incredible!"
www.youtube.com/watch?v=Sx0J...
March 21, 2025 at 5:28 PM
Interesting example of different intuitions about effect sizes.
@davidshor.bsky.social, who has run a lot of randomized experiments:
"You say the 70 words to him and then ... there's a 2.5% change he changed his mind. I think that's incredible!"
www.youtube.com/watch?v=Sx0J...
@davidshor.bsky.social, who has run a lot of randomized experiments:
"You say the 70 words to him and then ... there's a 2.5% change he changed his mind. I think that's incredible!"
www.youtube.com/watch?v=Sx0J...
The studies I know have been inconclusive. One of the more solid results is that swapping out algorithms, yes, reduces time spent but also news knowledge...
(Results from the 2020 Facebook/Instagram studies, published in Science & Nature)
(Results from the 2020 Facebook/Instagram studies, published in Science & Nature)
March 16, 2025 at 4:58 PM
The studies I know have been inconclusive. One of the more solid results is that swapping out algorithms, yes, reduces time spent but also news knowledge...
(Results from the 2020 Facebook/Instagram studies, published in Science & Nature)
(Results from the 2020 Facebook/Instagram studies, published in Science & Nature)
When you write this opening part, do you have specific evidence in mind re: polarization and how "easy" benevolent algorithms would be?
March 16, 2025 at 4:55 PM
When you write this opening part, do you have specific evidence in mind re: polarization and how "easy" benevolent algorithms would be?
I put this vintage issue of @wired.com out on my desk at MIT this week
March 15, 2025 at 4:03 PM
I put this vintage issue of @wired.com out on my desk at MIT this week
"The bottom 80% of earners spent 25% more than they did four years earlier, barely outpacing price increases of 21% over that period. The top 10% spent 58% more."
www.wsj.com/economy/cons...
www.wsj.com/economy/cons...
February 24, 2025 at 3:24 AM
"The bottom 80% of earners spent 25% more than they did four years earlier, barely outpacing price increases of 21% over that period. The top 10% spent 58% more."
www.wsj.com/economy/cons...
www.wsj.com/economy/cons...
I'll assume he's saying that some pre-2014 Russian territory may need to become an empty buffer zone
February 13, 2025 at 2:15 AM
I'll assume he's saying that some pre-2014 Russian territory may need to become an empty buffer zone
A few years back, I put together this t-shirt commenting on changing political norms
www.zazzle.com/peanut_farm_...
www.zazzle.com/peanut_farm_...
December 30, 2024 at 8:37 PM
A few years back, I put together this t-shirt commenting on changing political norms
www.zazzle.com/peanut_farm_...
www.zazzle.com/peanut_farm_...
Message at the supermarket today in Gualala
December 22, 2024 at 10:32 PM
Message at the supermarket today in Gualala
Sometimes survey questions just aren't really thought through, with likely consequences for results...
Example: congestion pricing for entering Manhattan. Where's a good answer for the 6.5M people in NYC ex-Manhattan, the vast majority of which don't usually Manhattan by private car.
Example: congestion pricing for entering Manhattan. Where's a good answer for the 6.5M people in NYC ex-Manhattan, the vast majority of which don't usually Manhattan by private car.
November 22, 2024 at 7:01 PM
Sometimes survey questions just aren't really thought through, with likely consequences for results...
Example: congestion pricing for entering Manhattan. Where's a good answer for the 6.5M people in NYC ex-Manhattan, the vast majority of which don't usually Manhattan by private car.
Example: congestion pricing for entering Manhattan. Where's a good answer for the 6.5M people in NYC ex-Manhattan, the vast majority of which don't usually Manhattan by private car.