The Economist Is Economical About Intelligence
The Economist is not a reliable guide to human ability. It believes in homo-economicus: all-purpose, equipotential beings without inherent differences, any such adventitious peculiarities to be smoothed away by compensatory education and the amelioration of unfavourable circumstances.
Very occasionally, like a maiden aunt reluctantly acknowledging the existence of sexual arousal, they refer to genetic differences, but soon revert to their standard mantra: with more education, earlier education, and more flexible education those nasty gaps between one person and another, and one group and another, can be washed away. Perhaps so.
As part of a holiday ritual I buy The Economist to take to the beach and find out if it has improved. The issue of June 25th looked promising, in that it had a special report on artificial intelligence, in which they said that the intelligent response to the dislocation caused by this development would be “making education and training flexible enough to teach new skills quickly and efficiently”. Quite so. They should have added “but the intelligent will always learn more quickly and will generalise their learning more widely, and to greater advantage”.
Learning speed is correlated with general ability. The US armed forces have all the data, and Linda Gottfredson has dug it up. The Wonderlic data also show that training a person in one simple task in one ability domain does not generalise to improved ability in simple tasks in other ability domains. A task is learned, but the individual is no brighter or faster at learning the next task.
They also champion social skills, which they say will be particularly required when robots strip out many humdrum jobs from the economy. However, social skills and character are largely personality characteristics which are heritable, and not easy to alter. (Heritability estimates for personality variable may be lessened by the poor reliability of self-assessed personality, so the use of better tests may show even higher heritability than the current 40％ or so). Some social skills can be trained, and this is currently very popular, because it is cast as “emotional intelligence” and everyone likes being intelligent in some way. Moving in a mysterious way, for example.
Measures of emotional intelligence turn out to contribute little to job success and response to training. Hunter and Schmidt.
Being one standard deviation above the norm leads to 60％ higher wages. These are OECD results, though they make absolutely no mention of intelligence.
The special report on artificial intelligence is worth reading. They point out that advances in technology usually create new jobs of different sorts, usually more complicated and demanding service jobs. Most people keep busy and are in work. The authors don’t mention that the current elite belief that mass immigration is required to provide low skilled workers does not sit well with the promises of artificial intelligence: governments seem to be hell-bent on bringing in people to do jobs which, if the advances of artificial intelligence are to be believed, shortly will not exist.
Although the report is interesting, one needs to fill in the empty spaces. It shows US employment by type of work (US Population Survey, Federal Reserve Bank of St Louis). The graph shows that the greatest increase since 1983 is in “non-routine cognitive” jobs, with routine cognitive and manual jobs stagnant, and even non-routine manual jobs increasing.
尽管这篇报道很有趣，但我们还是需要关注故事全貌。报道展现了美国不同种类工作的就业情况（美国人口调查，圣路易斯美联储银行）。图表显示，自1983年以来增量最大的是非例行性智力工作（Non-routine cognitive jobs），而在例行性智力工作（routine cognitive jobs）和例行性智力兼体力工作（Routine Cognitive and Manual jobs）停滞不变的同时，非例行性体力工作（Non-routine manual jobs）竟也有所增长。
They fail to point out that routine jobs are for those who need training on every step of a simple job, and that these people find departure from routine challenging, because those non-routine problems require higher ability. “Non-routine” jobs are where brighter people flourish because they do not need to be trained on each step of a task, but can grasp the general principles. This is increasingly the case at IQ 115 and above.
It is findings of this sort which make me question the predictions of Frey and Osborne (2013) as to which jobs are at risk of being replaced by artificial intelligence. They list many complicated non-routine jobs as being at risk: typists, technical writers, and accountants and auditors. Typists have already been reduced in number, but highly paid busy people still employ them, because on Ricardo’s principle of comparative advantage it is worthwhile for a highly skilled, highly paid individual to employ one so as to free up their own time.
In this way digital secretaries can proliferate in Scottish islands while executives fret in central London: these assistants are not strictly necessary but very useful. Equally, global job clearance centres (oDesk and others) can provide people to spruce up websites, enter data, draw graphs and generally tidy up stuff so that higher paid workers can get on with their higher paid enterprises. Will auditors become redundant? Seems unlikely. More likely that they will extend the scope of the feedback they give their clients. We shall see.
这样来看，数字秘书能够在苏格兰岛迅速扩张，而主管人员在伦敦市中心却不走俏，是因为数字秘书虽严格意义来说并非必需但却非常实用。同样地，全球各地的工作结算中心（job clearance center），比如oDesk及其他，能提供经营网站、输入数据、制作图表和广义上地“整理收拾东西”的人员，使得那些高薪工作者能继续干他们的“高薪事业”。审计员会失业吗？看上去不太可能。他们更可能会扩大提供给客户的服务范围。让我们拭目以待。
By the way, self-driving vehicles, all the rage at the moment, may turn out to be something of a waste of time, except perhaps in slow moving traffic jams. Incidentally, such jams might be better dealt with by road pricing systems: charge by the minute rather than the mile and road users will stagger their journeys or use satnavs and local knowledge to work round traffic jams.
The main point is that driving is too easy. Most people can do it. The better thing would be to automate motor maintenance and repair. Drivers are cheap, mechanics expensive.
Further into the magazine another story caught my eye. “A running start: Poor children fall behind early in life. Better pre-school education could help”. This notes with dismay that “By the time pupils begin primary school, there is a huge gap in achievement between rich and poor”. It further notes that the plus one standard deviation between 10th and 90th income percentiles in school achievement has barely diminished by age 18, equivalent to several extra years of secondary schooling.
How to explain this? Perhaps brighter parents earn more and have brighter children, with some drift downwards because of regression to the mean. Perhaps duller parents earn less and have duller children, with some drift upwards because of regression to the mean. This obvious genetic hypothesis is not mentioned. Instead (brace yourself) poor parents don’t talk to their children, because of poverty. “By the age of 6 children of wealthy parents have spent as much as 1,300 more hours in enriching activities than those of poorer families.”
How to test this, avoiding the blindly obvious confounder that the bright and wealthy parents are bringing up their own genetically brighter children? How about something more powerful than necessarily part-time pre-school activities? How about full-time adoption?
The article then goes into the Perry Pre-school project and a few selected Head Start programs. I sympathise. I could have written such an article 4 years ago, based on my favorite examples of early pre-school intervention, until I knew better. The general drift of the findings on pre-school intervention is disappointing. Interventions do not have lasting large effects on intelligence, and the welcome general beneficial effects are not often found, apart from the frequently mentioned star projects, which turn out to be outliers. Andrew Sabisky found the paper and the link is below:
文章接着提到佩里学前教育计划（Perry Pre-school project）和挑选的几个开端计划（Head Start programs）。我表示理解。四年前我也能拿自己最爱用的那几个早期学前干预当例子写出这样的文章，可随着我的了解逐渐加深，我发现学前干预的大体趋势令人失望。这类干预对智力并不能产生持续的明显效果，通常也无法实现那些为人喜闻乐见的整体效益，除了几个经常被提及的明星项目，但最后证明只是异常值而已。Andrew Sabisky找到了相关论文，链接如下：
Although the trend shows little effect, the outliers need further investigation. Ramey, talking at the ISIR conference in San Antonio, said that they reason he got results with the Abecedarian project whilst others did not, is that he made his teachers understand that if they didn’t get results they would not have a job. They followed protocols carefully, was his argument. He is running replication studies, this time with a full genome on each child, so we should have better results say 25 years from now.
There is nothing malign in offering compensatory education to young children, just significant doubt that it is effective and can be scaled up across entire education systems to smooth away real individual differences.
The Economist has massive reach among influential people in the world’s economies. The magazine, popular as it is with readers wanting to understand the ways that economies work, is resolutely behind the curve of contemporary research on human ability. One day The Economist may commission articles from people who know something about intelligence, education and genetics. For example, Ian Deary, Robert Plomin, Stuart Ritchie, and Tim Bates for a start. Until then, blogs like this one will be bringing the most recent findings to a small but highly select group: my dear readers. If your friends sometimes read The Economist, can you direct them to this post, in the hope that they can avoid the torrent of misunderstanding purveyed by the mainstream received wisdom? It would be a small mercy.
《经济学人》大量接触了全球经济领域里有影响力的人物。因为读者都想了解经济的运作方式，所以这本杂志非常受欢迎，但它明显跟不上目前关于人类能力方面的研究。有一天《经济学人》可能会去委托智力、教育和基因领域的专家写文章，比如从Ian Deary、Robert Plomin、Stuart Ritchie和Tim Bates开始。但在那一天到来之前，我这样的博客会一直把最新的研究发现带给精挑细选的一小部分人——也就是我亲爱的读者你们。如果你们有朋友在看《经济学人》，能给他们推荐这个博客吗？这样他们就可以避开主流思想带来的各种误解，也算是小小地行个善吧。
The Economist is confused about human beings, but if we are to accept their shaky presumptions about the power of pre-school education, then even journalists can be educated, if they can be caught young.
编辑：辉格@whigzhou comments powered by Disqus