Sean Harrison, PhD
@sean-h.bsky.social
People thinking chatbots are real has been a thing for, surprisingly, almost 60 years (1966), and received the name "The ELIZA effect" after the program: en.wikipedia.org/wiki/ELIZA_e...
![Quote from wikipedia: Weizenbaum first implemented ELIZA in his own SLIP list-processing language, where, depending upon the initial entries by the user, the illusion of human intelligence could appear, or be dispelled through several interchanges.[2] Some of ELIZA's responses were so convincing that Weizenbaum and several others have anecdotes of users becoming emotionally attached to the program, occasionally forgetting that they were conversing with a computer.[3][page needed] Weizenbaum's own secretary reportedly asked Weizenbaum to leave the room so that she and ELIZA could have a real conversation. Weizenbaum was surprised by this, later writing: "I had not realized ... that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people."[23]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:yv43iheuhokouafy55y4xjcb/bafkreigrxco4q5lcyp5b47edyafvkgkbpbep5i57jt5ypnjuijzoo4x5za@jpeg)
November 27, 2025 at 5:02 PM
People thinking chatbots are real has been a thing for, surprisingly, almost 60 years (1966), and received the name "The ELIZA effect" after the program: en.wikipedia.org/wiki/ELIZA_e...
The coefficient for sex will NOT be 0 if there is no direct effect of sex on final weight, because there is an indirect effect through true weight (unmeasured).
The coefficient for sex is the difference in constants in the statistician 2's results.
The coefficient for sex is the difference in constants in the statistician 2's results.
October 12, 2025 at 3:06 PM
The coefficient for sex will NOT be 0 if there is no direct effect of sex on final weight, because there is an indirect effect through true weight (unmeasured).
The coefficient for sex is the difference in constants in the statistician 2's results.
The coefficient for sex is the difference in constants in the statistician 2's results.
Essentially, I think you have a true value of weight, which is the underlying value of initial and final weight (weight probably doesn't change over time, but has measurement error).
Sex affects true weight, not initial or final weight (except through true weight).
True weight is unmeasured.
Sex affects true weight, not initial or final weight (except through true weight).
True weight is unmeasured.
October 12, 2025 at 3:06 PM
Essentially, I think you have a true value of weight, which is the underlying value of initial and final weight (weight probably doesn't change over time, but has measurement error).
Sex affects true weight, not initial or final weight (except through true weight).
True weight is unmeasured.
Sex affects true weight, not initial or final weight (except through true weight).
True weight is unmeasured.
Ed Byrne had a good bit a while ago about racism is to UKIP as vegetarianism is to the Green Party.
Seems to apply here.
Can't find it on youtube, but:
Seems to apply here.
Can't find it on youtube, but:
September 29, 2025 at 10:30 AM
Ed Byrne had a good bit a while ago about racism is to UKIP as vegetarianism is to the Green Party.
Seems to apply here.
Can't find it on youtube, but:
Seems to apply here.
Can't find it on youtube, but:
"Rabbits eat their own shit, and Bugs Bunny regularly outwitted Elmer Fudd. Given how similar in name 'Elmer Fudd' and 'Elon Musk' are, I decided to see whether eating my caecal products would increase my IQ across the Rubicon separating mortal man and technological deity: the median."
August 24, 2025 at 8:14 PM
"Rabbits eat their own shit, and Bugs Bunny regularly outwitted Elmer Fudd. Given how similar in name 'Elmer Fudd' and 'Elon Musk' are, I decided to see whether eating my caecal products would increase my IQ across the Rubicon separating mortal man and technological deity: the median."
Wait: The Starlings
August 14, 2025 at 12:11 PM
Wait: The Starlings
![Table showing country, legality of Nazi symbols, and exceptions.
For the United States: Legality: Restrictions in the states of New York, California and Virginia;[36][37][38]legal elsewhere; Exceptions: Unconstitutional to ban symbols outright. New York State ban only extends to public property. California bans Nazi symbols in public pursuant to the state's hate-crime law, and Virginia bans Nazi symbols in public to avoid confusion with the Hindu Swastika.
For the United Kingdom: Legality: Legal[35]; Exceptions: During World War II (1939–1945)
From WIkipedia: https://en.wikipedia.org/wiki/Bans_on_Nazi_symbols](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:yv43iheuhokouafy55y4xjcb/bafkreif3vl2bzcbvqnlcdnra3a7mq3csckfh4mmmhrvpxq3bklvr6326zi@jpeg)
I once did a systematic review and meta-analysis where one of the resulting (fixed effect) ORs was 1.00 (95% CI: 1.00 to 1.00).
I did *not* conclude that more research was needed.
If anything, less research was needed.
I did *not* conclude that more research was needed.
If anything, less research was needed.
August 11, 2025 at 8:25 AM
I once did a systematic review and meta-analysis where one of the resulting (fixed effect) ORs was 1.00 (95% CI: 1.00 to 1.00).
I did *not* conclude that more research was needed.
If anything, less research was needed.
I did *not* conclude that more research was needed.
If anything, less research was needed.
I thought the relative precision could be lower for Z because it was binary, and low number of imps could mean greater possibility of higher imprecision.
I re-ran with Z~N(0,1) - took out "if `run'", otherwise same code.
Now I just think "it was chance".
Also, why not 1.96*MCSE in graph?
I re-ran with Z~N(0,1) - took out "if `run'", otherwise same code.
Now I just think "it was chance".
Also, why not 1.96*MCSE in graph?
July 24, 2025 at 11:22 AM
I thought the relative precision could be lower for Z because it was binary, and low number of imps could mean greater possibility of higher imprecision.
I re-ran with Z~N(0,1) - took out "if `run'", otherwise same code.
Now I just think "it was chance".
Also, why not 1.96*MCSE in graph?
I re-ran with Z~N(0,1) - took out "if `run'", otherwise same code.
Now I just think "it was chance".
Also, why not 1.96*MCSE in graph?
I hadn't thought about whether DNA might be flammable or not, but turns out these guys had: phys.org/news/2013-03...
July 7, 2025 at 11:48 AM
I hadn't thought about whether DNA might be flammable or not, but turns out these guys had: phys.org/news/2013-03...
Charitably, I guess what they're looking for is someone who can break a seemingly complex problem down into component parts, each of which is fairly simple, then put it all together to get a coherent answer, presumably in a way they could then explain to other people, who would then also understand.
June 21, 2025 at 12:50 PM
Charitably, I guess what they're looking for is someone who can break a seemingly complex problem down into component parts, each of which is fairly simple, then put it all together to get a coherent answer, presumably in a way they could then explain to other people, who would then also understand.
All outcomes consistent, no obvious alternative reason I can see why the results show that exercise is more beneficial than health education.
Fewer recurrences, primary cancers, deaths from all causes...
I'd like to know *how* the randomisation was done (computer, phone line etc.), but whatever.
Fewer recurrences, primary cancers, deaths from all causes...
I'd like to know *how* the randomisation was done (computer, phone line etc.), but whatever.
June 2, 2025 at 11:12 AM
All outcomes consistent, no obvious alternative reason I can see why the results show that exercise is more beneficial than health education.
Fewer recurrences, primary cancers, deaths from all causes...
I'd like to know *how* the randomisation was done (computer, phone line etc.), but whatever.
Fewer recurrences, primary cancers, deaths from all causes...
I'd like to know *how* the randomisation was done (computer, phone line etc.), but whatever.
Read the methods and results, and it looks... Fine?
Headline aside, I didn't see anything immediately obviously wrong with the trial, which is really surprising given the results.
I initially thought, "they probably just did per protocol", but nope, ITT.
Headline aside, I didn't see anything immediately obviously wrong with the trial, which is really surprising given the results.
I initially thought, "they probably just did per protocol", but nope, ITT.
June 2, 2025 at 11:12 AM
Read the methods and results, and it looks... Fine?
Headline aside, I didn't see anything immediately obviously wrong with the trial, which is really surprising given the results.
I initially thought, "they probably just did per protocol", but nope, ITT.
Headline aside, I didn't see anything immediately obviously wrong with the trial, which is really surprising given the results.
I initially thought, "they probably just did per protocol", but nope, ITT.
This, no?
The outcome isn't the outcome from the RCT, it's the effect estimate from the RCT (how does the effect estimate change given some exposure).
BUT: there needs to be mechanistic evidence that the confounder IS a confounder, right?
Otherwise, it could be a moderator of the interaction.
The outcome isn't the outcome from the RCT, it's the effect estimate from the RCT (how does the effect estimate change given some exposure).
BUT: there needs to be mechanistic evidence that the confounder IS a confounder, right?
Otherwise, it could be a moderator of the interaction.
May 11, 2025 at 12:45 PM
This, no?
The outcome isn't the outcome from the RCT, it's the effect estimate from the RCT (how does the effect estimate change given some exposure).
BUT: there needs to be mechanistic evidence that the confounder IS a confounder, right?
Otherwise, it could be a moderator of the interaction.
The outcome isn't the outcome from the RCT, it's the effect estimate from the RCT (how does the effect estimate change given some exposure).
BUT: there needs to be mechanistic evidence that the confounder IS a confounder, right?
Otherwise, it could be a moderator of the interaction.
Where k = a scaling factor and c is a constant (could be k=1 and c=0 to give original equation).
Add ^i over the brackets if you need an exponent factor.
Add ^i over the brackets if you need an exponent factor.
May 5, 2025 at 12:57 PM
Where k = a scaling factor and c is a constant (could be k=1 and c=0 to give original equation).
Add ^i over the brackets if you need an exponent factor.
Add ^i over the brackets if you need an exponent factor.
I live by the principle that if you try all legitimate means to acquire something to read/watch/listen to etc., and all avenues are blocked, then copyright is acting against its own interests, and can be ignored.
In my naive view, this is "fair dealing" - not depriving anyone of revenue.
In my naive view, this is "fair dealing" - not depriving anyone of revenue.
May 3, 2025 at 8:53 PM
I live by the principle that if you try all legitimate means to acquire something to read/watch/listen to etc., and all avenues are blocked, then copyright is acting against its own interests, and can be ignored.
In my naive view, this is "fair dealing" - not depriving anyone of revenue.
In my naive view, this is "fair dealing" - not depriving anyone of revenue.
I like how even Green voters don't love Green voters
May 1, 2025 at 10:22 AM
I like how even Green voters don't love Green voters
Ok, but the firestation in Clevedon was the inspiration for the Pontypandy firestation in Fireman Sam.
April 3, 2025 at 10:43 AM
Ok, but the firestation in Clevedon was the inspiration for the Pontypandy firestation in Fireman Sam.
I know it must have been difficult to scale hobbits right in every scene, but... C'mon. It's hard to mistake those kids for Sam and Frodo...
March 11, 2025 at 12:35 PM
I know it must have been difficult to scale hobbits right in every scene, but... C'mon. It's hard to mistake those kids for Sam and Frodo...
In one simulation, I got up to a maximum of £2,097,152, meaning I lost 21 times in a row, and bet £4,194,303 in total. To win £1…
This graph shows the longest losing streak across each of the 10,000 simulations (remember, this is all to win £100 in total – that’s it):
12/n
This graph shows the longest losing streak across each of the 10,000 simulations (remember, this is all to win £100 in total – that’s it):
12/n
March 2, 2025 at 1:16 PM
In one simulation, I got up to a maximum of £2,097,152, meaning I lost 21 times in a row, and bet £4,194,303 in total. To win £1…
This graph shows the longest losing streak across each of the 10,000 simulations (remember, this is all to win £100 in total – that’s it):
12/n
This graph shows the longest losing streak across each of the 10,000 simulations (remember, this is all to win £100 in total – that’s it):
12/n
I mean, good for a phase 1 clinical trial, but these aren't results from an RCT, which is what it naively looks like with the 2 survival curves.
What would the % cancer free at 3 years have been for the 19 patients in the trial?
Without that, we can't know whether the treatment is effective.
What would the % cancer free at 3 years have been for the 19 patients in the trial?
Without that, we can't know whether the treatment is effective.
February 28, 2025 at 2:34 PM
I mean, good for a phase 1 clinical trial, but these aren't results from an RCT, which is what it naively looks like with the 2 survival curves.
What would the % cancer free at 3 years have been for the 19 patients in the trial?
Without that, we can't know whether the treatment is effective.
What would the % cancer free at 3 years have been for the 19 patients in the trial?
Without that, we can't know whether the treatment is effective.
The earliest reference making the claim is Bjarnason 1987: pubmed.ncbi.nlm.nih.gov/3609658/
Their claim was in the discussion, which said:
"inasmuch as >30 million patients receive NSAIDs on a regular basis worldwide"
There is no reference.
13/n
Their claim was in the discussion, which said:
"inasmuch as >30 million patients receive NSAIDs on a regular basis worldwide"
There is no reference.
13/n
February 22, 2025 at 8:33 PM
The earliest reference making the claim is Bjarnason 1987: pubmed.ncbi.nlm.nih.gov/3609658/
Their claim was in the discussion, which said:
"inasmuch as >30 million patients receive NSAIDs on a regular basis worldwide"
There is no reference.
13/n
Their claim was in the discussion, which said:
"inasmuch as >30 million patients receive NSAIDs on a regular basis worldwide"
There is no reference.
13/n
Anyway, I made a flowchart showing each of the journal articles and websites I found making the claim, and what (and if) they were referencing to support the claim:
Orange references don’t have the claim, blue do, and green are websites.
12/n
Orange references don’t have the claim, blue do, and green are websites.
12/n
February 22, 2025 at 8:33 PM
Anyway, I made a flowchart showing each of the journal articles and websites I found making the claim, and what (and if) they were referencing to support the claim:
Orange references don’t have the claim, blue do, and green are websites.
12/n
Orange references don’t have the claim, blue do, and green are websites.
12/n
But, instead of being able to concentrate on the history of NSAIDs, I became distracted by a common piece of information, repeated in a good proportion of the articles and websites I was reading through.
"30 million people take NSAIDs worldwide every day"
3/n
"30 million people take NSAIDs worldwide every day"
3/n
February 22, 2025 at 8:33 PM
But, instead of being able to concentrate on the history of NSAIDs, I became distracted by a common piece of information, repeated in a good proportion of the articles and websites I was reading through.
"30 million people take NSAIDs worldwide every day"
3/n
"30 million people take NSAIDs worldwide every day"
3/n