Category: Research

No, that doesn’t settle it.

A Washington Post article about a research paper claims to have settled the question of whether to put one space or two spaces after a full-stop (or as Americans like to say a ‘period’*).

The research neatly encapsulates some of the elements of questions of objectivity and meaning that I keep returning to.

The research had two components. The first was about usage and is interesting but not consequential. The second is of more note. Using eye-tracking, the researchers measured how a number of people followed sentences they were reading. Using that data, they could compare the relative reading ease of texts that used a single space after a full-stop and texts that used two spaces.

The results showed a small advantage for two spaces. By ‘small’ I mean:

  • ‘comprehension was not affected by punctuation spacing’ i.e. there was no measurable difference in how well subjects understood the texts they were reading.
  • there was some evidence that ‘initial processing of the text was facilitated when periods were followed by two spaces’.

So practically, two-spaces was not obviously better but MAYBE it required a smaller effort to read, perhaps. Note this second conclusion requires its own chain of inference that’s not well established i.e. it assumes that the processing of the text was facilitated but that was not measured directly.

But the bigger issue (mentioned in the WP article but not in the abstract of the paper) was that the text used was…

 ...in a monospaced font.

That does not make any of the findings in the report invalid. It doesn’t undermine the quality of the methodology used. It doesn’t make the findings less objective BUT it does entirely miss the point of the underlying argument.

The two-space versus one-space debate pertains to the transition from typewriters to modern wordprocessing. Classic typewriters had to use common widths between letters due to the mechanics of a typewriter, including a degree of error as to exactly where a letter might be placed. Modern word-processing uses typefaces where letters and the spacing around them are customised for not just individual letters but also for punctuation. The two-space versus one-space argument is one about the transition from classic typing to modern word-processing.

There is a parallel with drug trials here. For example a new drug or treatment might be compared with a placebo. That’s a scientifically legitimate approach to collecting data and looking at efficacy. However, its often not the relevant question. More pertinent is how the new drug compares with existing treatment rather than a placebo.

The point being – what is the underlying issue or what is the question being asked? These are more vague, more wooly aspects of scientific inquiry but also deeply important. The more clarity on those aspects help us judge whether empirical evidence is relevant to the question being asked.

However, my point above does not mean the research was wasted. It does demonstrate a couple of things:

  • The typing habit of using two spaces after a full stop had some merit.
  • The possible advantage of using two spaces is very small.

I don’t think either conclusion helps out the two-spacers much. The first implies social habits and vague aesthetics of people who type can be trusted – and that would tend towards favouring the one-spacer’s attitude to modern texts with modern fonts. The second implies that the cost-benefit of using two spaces is a best marginal and at worst a waste of time. Although, I’m clearly showing my one-spaced prejudices here.

*(As we are engaged in trivial quibbles of no actual consequence, let me just say that ‘period’ should be retired as a name for the full-stop. It should then be re-allocated to the n-dash whose role is often to indicate a period of time, such as when it joins two dates together. I also have opinions about hyphens and dashes that I will reserve for another post – I feel the controversy would just be WAY too much for you all.)

Advertisements

Objectivity and stuff

I wanted to write about some of the interesting things people have been saying about reviewing but part of my brain obviously wants to talk about reason and evidence and those sorts of things. I guess I haven’t done much of that this year in attempt to look less like a philosophy professor.

Anyway – objectivity! The thing with objectivity as a word is that we (including myself) use it in a way that implies various things which maybe aren’t really part of what it means. Objectivity carries positive connotations and connotations of authority in contrast to subjectivity. Those connotations suggest impartial judgement and a lack of bias. That’s all well and good – words can mean whatever a community of users want them to mean but I think it creates confusion.

Here is a different sense of ‘objective’ – to say something is objective is to say that two people can follow the same steps/process and come up with the same answer reliably. Maybe we should use a different word for that but such processes are often described as ‘objective’ because they clearly contrast with subjective judgement.

The thing is that meaning does not in ANYWAY imply a lack of bias. Lots of systematic or automated processes can contain bias. Indeed we expect there to be biases in, for example, processes for collecting data. More extreme examples include machine learning algorithms which are inherently repeatable and ‘objective’ in that sense (and the sense that they operate post-human judgement) that nonetheless repeat human prejudices because those prejudices exist in the data they were trained on.

Other examples include the data on gender disparity in compensation for Uber drivers – the algorithm was not derived from human prejudices but there was still a pay disparity that arose from different working patterns that arose from deep-seated social disparities.

However, there is still an advantage here in terms of making information and data gathered more objective. Biases may not be eliminated but they are easier to see, identify and quantify.

Flipping back to ‘subjective’, I have discussed before both the concept of intersubjectivity (shared consensus opinions and beliefs that are not easily changed) as well as the possibility of their being objective facts about subjective opinions (e.g. my opinion that Star Trek: Discovery was flawed is subjective but it is an objective fact about the universe that I held that opinion).

Lastly the objective aspect of data can be mistaken for the more subjective interpretation of the data. In particular the wider meaning or significance of a data set is not established simply by the fact that the data is collected reliably or repeatedly.

Consider another topic: IQ. I’ve discussed before aspect of IQ and IQ testing and much of the pseudoscientific nonsense talked about it. Look at these two claims between Roberta and Bob:

  • Roberta: My IQ is higher than Bob’s.
  • Roberta: I am more intelligent than Bob.

The first statement may be an objective fact – it is certainly a claim that can be tested and evaluated by prescribed methods. The second statement is more problematic: it relies on opinions about IQ and the nature of intelligence that are not well established. The objectivity of the first statement does not establish the objectivity of the second. Nor does the apparent objectivity of the first imply that it does not have biases that may also impact wider claims based upon it.

Reading Peterson 11 – Notes & Facts & Hypothesis

Part 1, Part 2, Part 3, Part 4, Part 5, Part 6, Part 7, Part 8, Part 9, Part 10, Part 11, Part 12,…

There’s no shortage of notes in Jordan B Peterson’s book 12 Rules for Life but that doesn’t mean every assertion related to facts is referenced. Also, when references are used they aren’t always tightly associated with the argument. Take this for example from chapter 2:

“This is perhaps because the primary hierarchical structure of human society is masculine, as it is among most animals, including the chimpanzees who are our closest genetic and, arguably, behavioural match. It is because men are and throughout history have been the builders of towns and cities, the engineers, stonemasons, bricklayers, and lumberjacks, the operators of heavy machinery.” – Peterson, Jordan B.. 12 Rules for Life: An Antidote to Chaos (p. 40). Penguin Books Ltd. Kindle Edition.

Now there is a lot wrong with that statement factually but the right reference here, if this was an academic essay, would be to a source discussing historical patterns of employment. Peterson instead links to some modern labour statistics here https://www.dol.gov/wb/stats/occ_gender_share_em_1020_txt.htm The tables do use the term ‘traditional occupations’ and ‘non-traditional’ based on proportions of women involves but this is ‘traditional’ in a very loose sense and includes “Meeting, convention, and event planners”. My point here isn’t that the table is wrong of even questioning gendered-roles in employment – just that a lot of references are weak in this fashion. It is vaguely related but not neatly tied to Peterson’s argument.

(This is quite long – so more after the fold)

Continue reading

You say ‘a-loomin-um’, I say ‘al-you-min-ee-um’, we both say ‘bunkum’

I resolved to not bother talking about Vox Day for awhile but circumstances compel me. The synergies of nonsense bind extreme nationalism, Trumpism, misogyny, creationism and antivaxxerism. It is always remarkable to see what apparently scientific studies the Alt-Right will quote as if gospel and which they will turn their selective scepticism too.

To wit: https://web.archive.org/web/20171202043719/http://voxday.blogspot.com.au/2017/11/a-new-meaning-to-metalhead.html

What is all this about? It is the old and thoroughly debunked canard that vaccines cause autism. The idea is rooted in two coincidences: an increase in the numbers of people diagnosed with autism (primarily due to better clinical descriptions of autism spectrum and increased awareness among doctors and the public) and the timing of when autisim symptoms are often identified at an age close to when early childhood vaccinations occur. Campaigners against vaccinations have been looking for a more substantial way of linking the two and one generic culprit has been ‘toxins’ in vaccines – i.e. various additives used in the manufacture of vaccines. For a long time the supposed guilty party was mercury, particularly in the form of thiomersal – a preservative used in some vaccines. However, studies linking the two were famously debunked and many vaccines didn’t use thiomersal or other mercury compounds anyway.

Of later the antivaxxers have been pointing their fingers at a different metal: aluminium – which is just like the metal aluminum but more British. ‘Aluminium adjuvants’  are an additive to vaccine that use aluminium. Adjuvants are any substances added to vaccines whose role is to provoke an immune response (see here for a better explanation https://www.cdc.gov/vaccinesafety/concerns/adjuvants.html ). Tiny amounts of aluminium are added intentionally because the body’s immune system will react to the aluminium and it is that principle (which is central to the whole idea of vaccines) that has vaccination critics concerned.

Back to the study quoted. Vox Day is quoting from The Daily Mail:

http://www.dailymail.co.uk/health/article-5133049/Aluminium-vaccines-cause-autism.html

BUT….the Mail article is little more than a cut and paste from here:

https://www.hippocraticpost.com/infection-disease/aluminium-and-autism/

Which is an article by a “Chris Exley” who mainly writes alarming articles about the terrible things aluminium might do to you. Exley  is quoting a study from Keele University which is available here:

http://www.sciencedirect.com/science/article/pii/S0946672X17308763

And that study was conducted by three people including…Professor Chris Exley. Who, conincidentally enough is on the editorial board of the journal the study is published in:

https://www.journals.elsevier.com/journal-of-trace-elements-in-medicine-and-biology/editorial-board

It is a long chain and yet oddly this is a rare case where the populist half-baked version of the study is alomost directly from the scientist involved.

Now I don’t know much about Professor Exley’s field, so I can’t really comment on the validity of the methods used. The study involved detecting aluminium in a very small number of samples of brain tissue from dead people who at some point in their lives had been disagnosed with an Autism Spectrum Disorder. There’s not much in the way of comparisons in the paper and I get the (perhaps mistaken) impression that the method is relatively new. The paper correctly concedes that “A limitation of our study is the small number of cases that were available to study and the limited availability of tissue.”

But take a critical look at the next step in the reasoning. Exley hedges what he says but Vox follows the dog whistle:

“So, the obvious question this raises is: how did so much aluminum get into the brain tissue in the first place? And the obvious answer is: from being injected with vaccines containing aluminum.” (Vox Day)

Of course a moments thought reveals that cannot be the answer. Most people do not have a diagnosed Austism Spectrum Disorder but most people are vaccinated. For Exley’s hypothesis to be correct there would need to be some additional factor, which Exley does describe in his media article:

“Perhaps there is something within the genetic make-up of specific individuals which predisposes them to accumulate and retain aluminium in their brain, as is similarly suggested for individuals with genetically passed-on Alzheimer’s disease.”

Well perhaps there is but Exley’s study doesn’t show that. More to the point, if this IS true then vaccines and aluminium adjuvants are irrelevant – we are encounter far more aluminium in our diets than we do from the tiny amounts we might get from vaccinations. Exley has zero reason to point at vaccines, indeed his speculation would imply that vaccines CANNOT be the main reason for larger amounts of aluminium in his samples because neccesarily bigger sources are more likely.

Exley appears to be trying to join two different healthscare bandwagons together: general concerns about aluminium in stuff (see his other posts) and antivaxxerism.

Is the study itself flawed? As I said, I don’t know but the connection the paper makes to vaccines has zero substance and no evidence from the study itself. That in itself should have raised red flags with reviewers.

In the past, I’d have gone to Science Blogs for some extra background on something like this but that venerable home of blogs has been wound down.

Luckily ‘Orac’ of Respectful Insolence has set up their own blog here https://respectfulinsolence.com/ and has a deep dive into Exley’s paper here:

https://respectfulinsolence.com/2017/11/29/christopher-exley-using-bad-science-to-demonize-aluminum-adjuvants-in-vaccines/

Yup, it is as dodgy as somebody dodging things in a dodgy dodge. Orac points out the dubious funding source:

“The second time, I noted that he’s one of a group of scientists funded by the Child Medical Safety Research Institute (CMSRI), which is a group funded by Claire and Al Dwoskin, who are as rabidly antivaccine as anyone I’ve seen, including even Mike Adams. Among that group of antivaccine “scientists” funded by CMSRI? Anthony Mawson, Christopher Shaw, Lucija Tomljenovic, and Yehuda Shoenfeld, antivaccine crank “scientists” all. And guess what? This study was funded by CMSRI, too. Fair’s fair. If antivaxers can go wild when a study is funded by a pharmaceutical company and reject it out of hand, I can point out that a study funded by an antivaccine “foundation” is deserving of more scrutiny and skepticism.”

And it just gets worse from there. No controls, some tiny sample jiggery-pokery with the numbers and so on. Best read directly.

 

Hulk Donuts

There was an excellent question on Twitter that caught my attention:

I think firstly this is not quite a science question. If you need reliable answers to questions about fictional things then mathematics is the place to turn…

But first a diversion into biology. The space that forms your alimentary canal isn’t really you. It is really a great big hole that goes through you. The stuff that goes in there gets mashed and melted and broken down but it is only as that stuff crosses over the membranes that separate you from your innards that the stuff becomes part of you. The stuff in your gut is no more you than the stuff on your skin or in your hair. The stuff in your stomach (etc.) is not you. Therefore, the stuff in Bruce Banner’s tummy (or the Hulk’s) isn’t Bruce Banner. Hold that thought.

Topology is a lovely branch of mathematics. It is a way of looking at spaces without getting too hung up about shape. For example we can stop thinking about how Bruce Banner looks different to The Hulk for a moment and instead consider them both more simply.

Bannerhulk-space is the space occupied by Bruce Banner, The Hulk or intermediary states between the two. When Bruce is Bruce, Bannerhulk-space is small. When the Hulk is the Hulk, Bannerhulk-space is at its largest. Make sense? OK, now flip that. What about Not-Bannerhulk-space? That is all the space in the universe that isn’t Bannerhulk space.

Now when Bruce is Bruce, Not-Bannerhulk space is at its largest. When the Hulk is the The Hulk, Not-Bannerhulk space is at its SMALLEST. Aha! Now what about The Hulk’s guts? That’s part of NOT-Bannerhulk space! That doesn’t prove anything but it does point to the fact that we should assume (as a starting point) that The Hulk’s guts are actually SMALLER than Bruce Banners.

Hmmm. Can I show this? I can! Topology again. A person is nearly, topologically, a torus i.e. a doughnut shape. People aren’t quite a torus because of our nose holes but we can ignore those (no, your genitals/urinary tract don’t count as holes topologically – sorry).

We can model the Bruce Banner-Hulk transition as a torus with changing radii. There are two to consider – the radius of the ring overall and the radius of the tube that forms the torus. Now if The Hulk was simply a scaled up version of Bruce Banner then there wouldn’t be an issue – and also the special effects in The Avengers would be much simpler (paint Mark Ruffalo green and have him stand closer to the camera). Instead The Hulk gets bulkier as well as bigger – the equivalent of both radii increasing but the radii of the tube increasing disproportionately.

Let’s see what happens:

hulkdonut

Now maybe you are thinking at this point that all of the earlier part of this essay was really just a rationalisation for the time I spent making that GIF…in which case…yeah, maybe.

Back to biology. When Bruce Banner turns into The Hulk, I think the mathematics shows that his alimentary canal will actual constrict. This will force the contents of Bruce Banner’s digestive system out through his orifices – possibly violently so. Now, I note this doesn’t happen in the comic books or movies but maybe it occurs off screen. Just saying.

Blogstrology

It is only a tiny step from pointless science to pseudoscience and I’m thinking…it’s a rainy Sunday and my head hurts…

After my previous post on this topic, it occurred to me that I should check the profile of some other websites. I’d already identified that Vox Day’s blog was disproportionately Goat-Wolf-Rabbit. What about Monster Hunter Nation?

mhiblogstrology

A clear Tiger-Goat-Cow blog. Cats do quite well at MHI in terms of raw numbers but not when compared against their general frequency.

Moving away from the right, how about File770?

770blogstrology

Mike is running a Cat-Tiger-Goat blog it seems. Now note that the search method includes comments, so it may be the readers that have a thing about cats (this has been independently confirmed).

What do all three blogs have in common? GOATS.

[ETA – Rocket Stack Rank www.rocketstackrank.com is interesting because the animals mentioned would be more determined by their incidence in short fiction. Overall low frequencies and RSR has no presence on the otter or goose dimensions. Wolf-Rabbit-Cat blog – “Cat” strongly assisted by reviews of the works of Cat Rambo 🙂

Goat has a presence but is just shy of the top 3.

RSRanimals

Today in Pointless Statistics

Yesterday, I was speculating about how the far-right may have a fear of rabbits. I’ve no means of ascertaining that but I did wonder if rabbits got mentioned more than you would expect.

Disproportionate Lagomorphic Referencing in Ideologically Extreme Propaganda

By C.Felapton, M.Robot 2017

Abstract

It has been postulated that the alt-right talks about rabbits a lot. Our research unit examined this hypothesis empirically using highly advanced data-mining techniques.

Using a sample of common animal words, the frequency of use of those words was established and then compared with word frequency in an established corpus of English words. It was established that at least one member of the alt-right talks about rabbits disproportionately.

Method

A weblog site produced by a notable “alt-right” writer was identified by a process of his being the obvious one to have a look at (a blogger who use the pseudonym of “Vox Day”). A set of 14 common animal nouns was identified: cat, chicken, cow, dog, elephant, goat, goose, mouse, otter, rabbit, sheep, tiger, wolf.

For comparison purposes, a corpus of English words was identified to establish standard frequencies for each word. The selected corpus was the BYU-BNC.

The British National Corpus (BNC) was originally created by Oxford University press in the 1980s – early 1990s, and it contains 100 million words of text texts from a wide range of genres (e.g. spoken, fiction, magazines, newspapers, and academic). https://corpus.byu.edu/bnc/

Using Google’s site specific search function, the target website was searched using each animal word in turn as the search term. An example search query being “mouse site::http://voxday.blogspot.com”

The number of “hits” per search term was recorded.

Results

The most common animal name used from the sample was “dog”. However, given the very high frequency of “dog” in English, this result is unremarkable. The ratio of the blog frequency versus the corpus frequency was calculated. The mean ratio for the sample was 0.728 (to 3 s.f.) [blog freq/BNC freq].

The most disproportionately under mentioned animal was “mouse”. The most disproportionately over mentioned animal was “goat”. While the frequencies of both “rabbit” and “wolf” were quite different in both the blog and the corpus, both words were over mentioned in a similar ratio (1.20 for rabbit and 1.21 for wolf).

Full results are shown in Appendix A.

Discussion

It was agreed by the research team that this had been a pointless exercise that provided no valuable insights and which was methodologically flawed due to its arbitrary choice of words, blog and corpus. Meat Robot complained about having a cold a lot and suggested that a day spent re-watching Rogue One: A Star Wars Story would be a better plan. “You’re not the boss of me.” said Camestros but had to concede that it was impossible to exist as incorporeal being.

A cat refused to comment on the result and no other animals were consulted.

Appendix A: Full results

The table shows the full results in ascending order of ratio.

Animal Blog Freq BNC Freq Ratio
mouse

559

1,728

0.323

chicken

757

2,027

0.373

otter

79

188

0.420

sheep

1,420

2,942

0.483

cow

662

1,334

0.496

cat

1,880

3,788

0.496

dog

4,880

7,780

0.627

elephant

616

892

0.691

goose

404

479

0.843

tiger

764

870

0.878

rabbit

1,670

1,393

1.199

wolf

975

804

1.213

goat

844

593

1.423