Why Dont You Blog?

Now, I am sure every one knows that New Scientist is “pop science” - scientific news processed for laymen. In general this is great as is gives people an insight into the wonders of science without the tedium of years studying. Peer review is great, but only in its place. New Scientist is not the place.

Sometimes, this causes problems.

In this weeks issue, there is an article titled “Read my lips… and my voice, and my face” (online version titled “Software spots the spin in political speeches“) which is (at best) bad science being used for electioneering purposes. On the surface this is nothing more than the old idea that you can tell when people are lying by their gestures and use of language. This is a subject close to my heart and generally falls foul of the greatest of problems - it is sort of true. Body language, eye access, word selection and the like can give you an indicator of lies (for example) but only in the broader context of the persons behaviour.

Take the often cited example of people rubbing their nose when they lie. Yes, some people do this. But most of the time it means the person has an itchy nose and nothing else. The same with eye-access (as highlighted in The Negotiator), but the problem is people are different - not everyone looks the exact same way. Language choice is possibly the worst indicator as this is dictated by your background, education and the like. Simply put, there is no easy way you can use this information as a reliable indicator of deception or misdirection. You need to study the person in a variety of controlled circumstances and build up a pattern of their behaviour.

With this in mind, we can return to the New Scientist article. It seems someone has come up with an automated way of monitoring the terminology used, the voice and the facial expressions of politicians to measure how much “spin” there is in their speeches. Amazingly this has not resulted in 100% returns each time. This is how it is described:

The algorithm counts usage of first person nouns - “I” tends to indicate less spin than “we”, for example. It also searches out phrases that offer qualifications or clarifications of more general statements, since speeches that contain few such amendments tend to be high on spin. Finally, increased rates of action verbs such as “go” and “going”, and negatively charged words, such as “hate” and “enemy”, also indicate greater levels of spin. Skillicorn had his software tackle a database of 150 speeches from politicians involved in the 2008 US election race (see diagram).

Now, this strikes me as inherently flawed given that politicians have their speeches written for them by teams of “experts” (who are more than capable of concluding which words mean which things), and are nearly always well coached in delivering them in a manner to “stir” the audience. It strikes me that adding an arbitrary judgement as to what is, or is not, spin gives nothing that even resembles science. In an attempt to dismiss this, Skillicorn (the systems creator) says:

Additionally, [Skillicorn] says, little details count: pronouns such as “we” and “I” are often substituted subconsciously, no matter what is written in the script.

But you have no idea which ones are added by the script writers, which ones are subconcious and you certainly still haven’t proven that using “we” means there is a lot of “spin” in the speech. We still don’t really know what “spin” is - is it a good or bad thing?

The “Headline” results of this study are that Obama’s campaign has more spin than any of the other politicians (+6.7, where 0 is average for a politician) while McCain’s campaign had the lowest (-7.58). It states this supports McCain’s claim to being a “straight talker” (*cough*) and on the surface looks like it is a Republican Political Campaign masquerading as Science. In the articles defence, there is some balance:

So the analysis appears to back up McCain’s claim that he is a “straight talker”. However, for the purposes of political speech-making this may not be an entirely good thing for him. “Obama uses spin in his speeches very well,” says Skillicorn. For example, Obama’s spin level skyrockets when facing problems in the press, such as when Jeremiah Wright, the reverend of his former church, made controversial comments to the press.

Great from a science point of view. We would like to think that the readers of New Scientist are able to accept the idea that spin is a positive force for a politican.

However (and this is supported by a quick scan of the printed media that have picked up on this), the general population are not. We have been indoctrinated by decades of thinking politicians spin is an inherently bad thing. This article has generated several headlines in the free media about Obama being full of spin and McCain being straight talking. Both can translate into political capital. Shame on the New Scientist.

One funny bit which never quite made it into the free-papers is this nugget:

“The voice analysis profile for McCain looks very much like someone who is clinically depressed,” says Pollermann, a psychologist who uses voice analysis software in her work with patients. Previous research on mirror neurons has shown that listening to depressed voices can make others feel depressed themselves, she says.

Well, that pretty much summed up the effect his speeches have on me.

It is during the US Presidential elections that I thank Loki I live in the UK….

Popularity: 20% [?]

Popularity: 20% [?]

There’s a good article on Pure Pedantry. It illustrates why any amount of classy social statistics manipulation can be pointless without some social science understanding.

The Pure Pedantry post is about the claim that men and women have different numbers of sexual partners. The blogger sees this as inherently unlikely and refers to a mathematician David Gale who put that very point of view in the New York Times.

The obvious conclusion - that people lie about these things and they lie in ways that are not random but reflect social values about gender - convinces the blogger.

However, he updates the post with an explanation of how the seemingly irrational results could come from a confusion over the meaning of averages and medians in the statistics. This is all true, of course. However, it is irrelevant.

Whatever measurement you choose, the results are bull. They can only ever be bull, because there is no society on earth where ideas of maleness and femaleness are not socially constructed. So the results might tell you something about the meanings we attach to being a man or a woman but cannot say anything reliable about the actions of real men and women.

A thought experiment. Ask a thousand anonymous man and women about the number of sexual partners they have had. Ask another thousand anonymous men and women about the number of partners they believe is “normal”. You could probably guarantee that the results will broadly match, with the variation that men will report themselves at one or two above the mark and women will claim to be a mark or two beneath it.

Another thought experiment. Ask a thousand people some anonymous factual questions about their driving - “How many times have you ever driven over the speed limit?” “Do you ever drive when you are drunk?” and so on. The results would make you wonder how there are ever any accidents. Very few people would report themselves as driving in a potentially lethal way. Anonymity wouldn’t in itself make the results true, would it?

And it’s not just bragging or minimising. People don’t even tell themselves the truth. It’s the nature of society. Scientists aren’t immune from social values. A commenter on Pure Pedantry talks about girls with high numbers being “easy.” There is an entire lifetime of social conditioning in that phrase.

It’s a good post about how the choice of statistical methods can affect the interpretation of survey data. But there are some aspects of humanity that present far too much of a challenge to empirical observation to produce any meaningful results without some understanding of semiotics, cultural values, gender power relations, individual psychology and the whole rich world of social science knowledge.

I’m certainly not denying the validity of empirical data in the social sciences. The more the better. It’s just that we are nowhere near the point at which we can use people’s responses to questionnaires as evidence about their behaviour in the same direct way that we can use the number of bacteria on a slide as evidence for the presence of foot and mouth disease.

Popularity: 19% [?]

Popularity: 19% [?]

Thinking it might be a good idea to find out who Dorothy Rowe is (see last post) in case I was giving props to some spurious writer, I did the standard Google and found her website.

She is apparently a clinical psychologist who has published loads on depression. Her words seem to get mentioned as holy writ in lots of blogs, as far as I can see from Technorati

Dorothy Rowe Named As One of the Six Wisest People in the UK

This was apparently in a poll for Saga magazine for which she writes. Or wrote, because according to a page on her website, she’s just got the push by the very people who ran the poll which put her in the wisdom top 6.

I must say I had never heard of her (maybe I’ve just never been miserable enough) but she seems like everyone’s ideal mother or grandmother or great grandmother.

Respect. You’re better off writing for New Scientist anyway.

Popularity: 21% [?]

Popularity: 21% [?]

Again, more from the weird web department. This time, stumbling around the net brought me to a web page titled “How to spot a liar.” This is a page which explains how you can use eye movements, verbal constructions and blood flow to spot when people a lying. It is all packaged together well, and is generally an easy to read site.

The problem is, it is nonsense.

The bigger problem is that it is not pure, obvious nonsense, but the insidious nonsense which is latched on to some truth and then muddies the waters. Basically put the site discusses how eye movements can show which parts of the brain are being used, and how these parts of the brain have been (broadly) mapped onto construction or recall. That is about where it ends it’s relationship with reality.

For example, the site claims that when you ask some one a question, and they answer following a rapid eye movement up and to their right (your left), this means they are “constructing” the response and therefore lying. If they look up and left (your right) they are recalling the response and therefore telling the truth. As I said before, this is generally correct, but nearly half the population have this reversed. Makes using this a bit of a problem unless you know what you are looking for as you have an almost even chance of getting it wrong…

Add in to this the hazy use of recall and construct when it comes to answering a question and you can see that the most likely effect of taking this site seriously is to make you never know when some one is lying to you. Part of the art of getting a feel for deceit using clues like this, is learning how the question you ask influences the answer. Without that, even if you know which way the person looks, it wont help.

The examples given on the site are useful in this, and they highlight how the author of this post is turning slightly imperfect knowledge into a bad conclusion. This is the first example the author uses:

Popularity: 33% [?]

Popularity: 33% [?]

In recent weeks, any science content in New Scientist seems to be purely coincidental, with more and more pages being given over to woo and thinly veiled mysticism. This weeks issue is a minor deviation from this pattern, although most of the “solid science” is to be found in the letters pages…

There is one article, in the Comment and Analysis, which I am unsure about. Reading it, triggers a “bad science” response in me, but I am aware this may be a bit hasty. In an article titled “The media make a killing,” Michael Bond looks at some of the issues around the coverage of the Virginia Tech shooting. This is a well written article, which carries a lot of the “self evident truths” which the print media seem to like. As I was reading it, though, a few alarm bells were triggered — but this is not a subject in which I am well versed so before I scream Bad Science, I would like second opinions.

Popularity: 30% [?]

Popularity: 30% [?]