Mandela was right: the Foreign Language Effect
mappingignorance.org
mappingignorance.org
It's disappointing. We all know these kinds of polls can be easily gamed by subtle (or not-so-subtle) phrasing, and without the questions actually used, I have no way of knowing if the "Foreign Language Effect" actually exists or if their choice of words in the translation affected the outcome.
[1] http://psychology.uchicago.edu/people/faculty/foreignLanguae...
There's a sampling bias here. A similar dynamic goes on with Physics.
But, I'm sure that your general opinion is well-informed as well, or you wouldn't make such strong claims about a very broad field. What's your background in psychology, and what makes you say the things you do about it?
I don't believe you are being genuine here at all. This passive-aggressive, sarcastic response is not appreciated. If you don't like what I wrote, respond directly and honestly.
Nobel prize winner Daniel Kahneman[1] would disagree with you, along with social psychologist Jonathan Haidt[2].
[1] http://www.nature.com/news/nobel-laureate-challenges-psychol...
And this isn't exactly a new problem for psychology. Feynman somewhat famously questioned a lot of classic research with http://neurotheory.columbia.edu/~ken/cargo_cult.html. I say somewhat famously because people in areas like physics have all read that essay. But in my experience psychologists are a little less inclined to read the criticism.
Even at the best of times, replication of results in a complex system is hard. See, for instance, http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fj... for evidence that most published medical research is wrong.
But with psychology it is even worse. We lump people together by symptom, not by root cause. It is like lumping together people with migraines, head injuries, and caffeine headaches together. Then you find out that coffee beats a placebo for helping headaches, and give them all coffee!
You think I'm exaggerating? Read http://www.nimh.nih.gov/about/director/2013/transforming-dia... for verification that the NIH considers it a high priority research item to come up with classifications based on root causes, so that we have a chance of even being able to do useful research!
Yes, psychology has a lot of well-meaning people, and they stumble across a lot of interesting stuff. But it should be read with a critical eye, because the field as a whole does not perform reliably enough to compel our trust.
This happens (in medicine, not just psychology), and there's definitely room for hypothetical improvement. But in many cases, it's hard to blame people, because there's no way to do better without doing an unknown but gargantuan amount of research. For example, lupus is (or was when I had the discussion with my mother) defined as exhibiting X of a list of Y symptoms, where Y > X. This leads to situations like the following:
1. You, the patient, present with X-1 symptoms of lupus. By definition, you don't have lupus, and treatment for lupus is not warranted.
2. Time passes.
3. You, the patient, come back to your doctor with one additional symptom of lupus. Medical doctrine now states that you have lupus (not so interesting) and that, more interestingly, those X-1 earlier symptoms are symptoms of lupus (which you didn't have when you developed those symptoms), not of some other yet-undiagnosed problem. You can now be treated for lupus.
Obviously, that satisfies no one and is crying out for a more "reality-based" definition. The problem is that no one has a candidate for the actual physical causes of lupus. The best anyone's ever been able to do is recognize that the same set of treatments are broadly effective against a constellation of symptoms, which may seem to be related to each other or not, but which co-occur to some degree in people who respond to these treatments.
Of course it's a high priority to learn what the root causes of symptoms are, but that doesn't mean it's easy or even, in the general case, possible. You have to start somewhere.
This is basically the same as publishing a psych paper with a Methods section including a line that says "and then magic occurs". It's like an unsafe cast in a programming language -- you totally escape the guarantees provided by the wider structure of the academic system.
Academia in general has serious problems. There's a fundamental problem of incentive, and until that's solved, nothing's really going to change. But psychology is definitely unfairly maligned.
(It doesn't help that plenty of people doing doing what's almost psuedoscience are still harbored in a lot of psych departments, but their methods are still more rigorous than HCI research in most cases.)
If you look at computer science papers in machine learning, systems, human-computer interaction, networking, robotics, or architecture, you'll almost always see an experiment of some sort. These are typically benchmarks, but there are plenty of other examples of empiricism in computer science.
But the experiments in computer science are horribly designed! They lack almost any control, and use statistical methods that would make psychologists wince. Hypothesis testing is rare to the point of nonexistance. (Psychology sometimes adjusts too far in the other direction, using complex analyses where a simpler one would do.)
Further, experiments are nearly never replicable, at all, because the systems evaluated are never made public, or the experimental environment is never recorded. For a great treatment of this trend in systems, see http://www-plan.cs.colorado.edu/diwan/asplos09.pdf. Pointing out a systematic artifact in that many experimental designs would be career-making in psychology, or any other science! But you can look at modern papers and see exactly the same mistake being made.
An average person, talks to himself multiple times in a day, evaluating decisions, simulating how things are/should be, thinking about course of action etc etc.
This is done in a language, which often ends up as mother tongue to the outside world.
But the missing link here is the mental stamina. The brain has a native locale language, but if the information is coming in any other language, which needs to be understood (decoded) first into the locale language, it consumes stamina & time.
But in the case of it being actually in the language in brain's native language, less time & stamina would be consumed in the same conversation, and more time and stamina would be left for actual thinking/analysis/evaluation/etc.. leading to better understanding between people.
Peace!
The idea that there is a "mother tongue" of your thoughts is a common misconception. You can consciously think about speech and language, but it's an exception to how your brain usually operates.
> This is done in a language
Why do you believe this?
This is a theory, but most bi/tri-langual people I talked to experienced some kind of shift in their way of modeling their thought, sometimes relying on a concept only existing in a specific language (that's what makes the reliance on the language obvious).
Basically they've demonstrated that if people spend a lot more time on a question they answer less emotionally (which seems like it would be true independent of the language).
If they can demonstrate the effect still holds with 2nd language speakers with native level proficiency then this would be a lot more interesting result.
Although I know there is a one-to-one mapping between meters and feet, I only grok a measurement given in meters.
There's probably some conditioning behind this: every day you are reminded that a pack of salami is 200g, while a pack of flour is 1kg, and so on ... so that for each measurement you just access your library of memories that relate the number to some previous experience.
Since a similar dictionary does not exist for the foreign units you learned in an artificial setting, and since your brain does not auto-update all your past memories based on the new info that 1m=3+3/8ft, you are left with words that are all understanding and no emotion.
I still don't get cups or the fluid measurements though, except that 12 oz a beer bottle and 16 oz is a medium soda bottle (medium in European terms).
Maybe this will help with the fluids. Looking at a table of fluid measurements, it is easy to see it is just a base 2 system.
2^-1 oz = .5 oz = 1 tablespoon
2^0 oz = 1 oz = 1 oz (2 tablespoons)
2^1 oz = 2 oz = 1/4 cup (4 tablespoons)
2^2 oz = 4 oz = 1/2 cup (8 tablespoons)
2^3 oz = 8 oz = 1 cup (16 tablespoons)
2^4 oz = 16 oz = 1 pint (2 cups)
2^5 oz = 32 oz = 1 quart (4 cups)
2^6 oz = 62 oz = 1/2 gallon (8 cups)
2^7 oz = 128oz = 1 gallon (16 cups)
I think it is quite beautiful in a way. That doesn't explain why we have 12 oz beers though.
I know the feeling.
That in the case of medecine B, there's a 66% that the result (600,000 people dying) is equivalent to the result of not taking any medicine.
Whereas when the problem is framed in term of gain, it's quite clear by the phrasing "66.6% chance that no one will be saved.".
That's the problem with this kind of questionnaire, and it's compounded by the fact that they do not provide the Korean/Japanese/French version of the questionnaire they used.
The article's commentary on the _Asian Disease_ problem which states "the number of certain deaths is the same in both versions of the problem.." is incorrect. In case A, the number of certain deaths is 400,000, in case B it's 0.
Or am I reading it incorrectly ?
My interpretation was that in the first situation, there is a possibility that everyone dies if you choose medicine A, therefore B was the right choice. Similarly, in the second situation, there is a possibility that everyone lives if you choose medicine A, therefore A was the right choice. Maybe I'm misunderstanding but there does not appear to be enough information in the question to determine the exact probabilities.
Medicine A will always save 200,000 people, with 400,000 dying.
Medicine B will save all 600,000 people 33.3% of the time, but everyone will die 66% of the time.
The expected number of deaths (the 'expected value' in game theory) is the same for both cases (200,000). Medicine A is easy to calculate (200,000 * 1.00 = 200,000). For Medicine B, the calculation is 600,000 * 0.33 + 0 * 0.66 = 200,000.
The idea is that the average number of deaths is the same for both cases; you are just trading the certainty of Medicine A (where you know you will save exactly 200,000) vs the chance to save everyone, but the risk of losing everyone.
"If you choose Medicine A, 400,000 people will die."
Tells you that 400k people will die if you choose medicine A. It says nothing about what will happen to the remaining 200k. I believe it is illogical to assume that the sentence implies the remaining 200k live. Note that if you were to ask me in real life, I'd probably be happy to make that assumption but that doesn't change the fact that it is an assumption based on missing information.
When you follow that by saying "If you choose Medicine A, 400,000 people will die", it is not illogical to assume that they don't mean "oh yeah, the other 200k will also die, too" This isn't a Mitch Hedberg joke.
In these scenarios, we are expected to make many basic assumptions that aren't included in the scenario. We are safe to assume that there isn't an unmentioned side effect of Medicine B that will leave everyone paralyzed, or that Medicine A won't turn the survivors into zombies.
His comedy was very hit-and-miss, but some of it was spectacularly funny.
I'm sure this form of problem analysis gives psychological experimenters fits.
see for eg:
And here is the problem framed in terms of gains:
If you choose Medicine A, 200,000 people will be saved. If you choose Medicine B, there is a 33.3% chance that 600,000 people will be saved and a 66.6% chance that no one will be saved.
Which medicine do you choose?
Although the number of certain deaths is the same in both versions of the problem, people take the safer option
The lack of scale prevents an analysis of proportionality.
For example:
600k out of 500,000,000? or 600k out of 1,000,000 or 1,000,000,000?
Given the downsides are equal, the omission begs the qeustion about the rate of success. The rate of success is not a trivial source of people's bias, because empirically "non-success" is often accompanied by adverse effects (a/k/a side effects) as well as additional cost. The result is a leading question. In other words, the language is manipulative.
But do we really need this study to tell us this? C'mon, that is stupid. And has nothing to do with Mandela.
The entire article is just a layer over this showing the supposed effects of asking the question in a foreign language (other comments have explained the likely problems of the experiment), and then another layer over that wrapping it in "this is totally relevant to you because Madiba".
I mean, yeah. Mandela was right. But not because of the foreign language effect.
The problem is that the rational answer to this depends on a huge amount of background information, eg, the value of certainly having some money, versus the benefit of the additional money.
Further, they account for stuff like that by using large sample sizes. Different people are choosing both options. It's just that the proportions hold despite this.
They've spent a good 30 years confirming this. If you really want to criticize it, it'd be worth putting together a list of studies investigating this and analyzing their methodologies for problems.
http://en.wikipedia.org/wiki/Framing_(social_sciences)#Frami...
B: 1/3 probability of 600k cures, 2/3 probability of 0 cures; expected value = 200k survivors
Of course, as software writers, we have probably been trained to analyze algorithms for expected performance and for the best and worst cases. Real-time programmers should always choose option A, as it has the best worst-case performance. Some programmers will choose A because it has lower computational complexity. Others will choose option B because it has better best-case performance, and in the worst case is no worse than doing nothing at all.
The scenario is supposed to be mathematically ambiguous, but the people who designed it obviously did not consider factors beyond just the average case and whether the numbers are framed positively or negatively.
You're reading it incorrectly. It's saying that the number of certain deaths is the same, whether it's phrased as "400,000 deaths" or "saving 200,000 people".
Millions more Japanese would have died defending their homeland if there were a land invasion.
To gain a benefit from this scenario, you would need to prove that "learning a language for X years" increases the chances of "listener response is more emotional".
This, however, is much more unlikely since accent and language proficiency come into play.
So no HN-style hack yourself goal can be derived (which I kinda hoped to find here)
Wasn't Latin a great invention? It was foreign to everybody (I mean after the fall of Rome) - looks like ideal for a language of diplomacy and science.
Very good.
What I'm getting at here is that this idea of Mandela's also applies to samewise monolingual speakers — as we each have our own idiolect, so far as descriptivist linguists are concerned.
Double very good.
Apartheid should have ended without violence.
Mandela was a radical communist who committed 156 acts of terrorism and never denounced violence. The economy and currency of South Africa crashed under Mandela. Unemployment went up and income went down. Apartheid was a terrible system but more people have been murdered every year since Mandela's presidency than before. There's more but you get the idea.