There are a number of very good software-driven language courses out there, which I've gone some way toward playing with, but not to the point where I can say I know what they all do. I should emphasize that, so far as I can tell, they're all perfectly good as teaching tools. I don't think that there are huge failings in them. However, I do think that there are some general problems that they don't usually acknowledge.
No matter how much the marketing literature insists that their system is either time-tested or else the result of new, innovative thinking, the success rate in teaching languages seems to be low. There are many postings in various language forums from people who find some or most systems to be lacking. There is certainly none that seems to stand out as effective in all cases.
Part of this may be motivation. People may buy a system expecting quick results, only to find that they are committing to more time and effort than they had anticipated. Motivation seems to be one of the most difficult-to-measure aspects of language acquisition, and yet one of the most important to ensuring success.
One of the best means of motivating a student is to demonstrate improvement in performance. Language instruction pioneer, James Asher describes the contrast effect, in which students initially see great progress in language skills, only to become discouraged when the initial rate of learning doesn't continue. I suspect this is even more pronounced when using computer software than it would be learning through books or tapes. Use of computer software tends to be repetitive and requires a fair amount of attention. Language acquisition seems to require concentrated effort at some times, with casual experience at other times. Tapes, or movies, or television are better at providing this than any of the software I've seen so far. (Not to mention actually speaking with someone in the language, of course.)
There are, of course, other ways to motivate the student. At times, I wonder whether this might not be the brilliant idea behind what I otherwise consider the most annoying feature of Rosetta Stone software - it's price. If someone has spent $300-$500 on a yellow box containing some text, software CDs, and a microphone, they have a strong incentive not to waste the investment.
Friday, October 30, 2009
Thursday, October 29, 2009
Non-language-related observations about OOPSLA
1.) SPLASH! OK, so I always know that "OOPSLA" was a really goofy name, and the emphasis on object orientation was wearing a little thin, but did someone insist on a Daryl Hannah movie theme when they came up with the new name? And if so, is it too late to lobby for "Clan of the Cave Bear Conference"?
2.) Next year in Reno! I get to save on air fare! I get to see whether I can repeat my ultra-nerd cred by hanging out in a casino all week and doing only tech stuff without gambling, like I did at the last HP Universe in Las Vegas.
3.) The best quote of the conference will have people Googling for "Bjarne Stroustrup" and "Sex", which should come up with his latest bon mot, or else lead you to the most disturbing nerd slash fiction ever. Also waiting for a transcript of the Liskov talk.
2.) Next year in Reno! I get to save on air fare! I get to see whether I can repeat my ultra-nerd cred by hanging out in a casino all week and doing only tech stuff without gambling, like I did at the last HP Universe in Las Vegas.
3.) The best quote of the conference will have people Googling for "Bjarne Stroustrup" and "Sex", which should come up with his latest bon mot, or else lead you to the most disturbing nerd slash fiction ever. Also waiting for a transcript of the Liskov talk.
Friday, October 23, 2009
Why Learning a Language Is Not Like Learning Math
My interest in language learning software grew out of my sense that most of the software I'd been playing with while trying to learn Chinese was doing little more than presenting images and sounds. As a software designer, I disliked the fact that the computer was being used as little more than a glorified tape recorder and slide show projector. This led to my first concept I started floating with people. Why couldn't language software do a better job tracking a student's progress and respond accordingly? Unfortunately, that's not a purely rhetorical question. Here's why it can't, or at least why it can't do so focusing exclusively on instruction.
Prerequisites
Anyone who went through high school understands prerequisites. To take calculus, you first have to pass trigonometry. To study Chemistry 102 you first have to pass Chemistry 101 or an equivalency exam.
The model can even be applied at a smaller level. Before studying combinatorial methods, you first need to demonstrate an understanding of set theory. The order of prerequisites suggests a fairly obvious order for the curriculum.
eLearning
This kind of analysis is used in some e-learning software. A colleague working at one such company described how their software tracks a grade school student's performance in various subjects. If the student is struggling in a particular subject, the software predicts what future subjects the student will have trouble in, and where the student can most effectively improve. This could help prioritize efforts to help that student. So for example, the fact that Johnny has difficulty spelling the word "thorough" might be less significant than his problems with memorizing multiples of twelve.
At first glance, this would seem to be applicable to language learning. I certainly was interested in the idea, given my background in workflow and business process analysis. When my colleague described his company's software, I felt I must be on to something in thinking it could be applied to language. As the student learned new grammar and vocabulary, the software could monitor progress and review those subjects that were lacking. For example, if a student of English was having difficulty with irregular past tense (such as "catch/caught"), the system could review the regular past and highlight the irregular exceptions. Unfortunately, there's one problem with this idea. Language learning doesn't work that way.
Acquisition vs. Learning
First off, the way that people learn a language is significantly different from how we learn other things. It would not be unusual for someone living in a foreign country to pick up at least some skill in the local language just through exposure to it. By comparison, it would be extraordinary for someone to gain insights in probability just by hanging out with mathematicians on their days off, but never attending classes.
Secondly, measuring progress in language is not necessarily the same as measuring progress in a subject like math. It is not uncommon for a language student to start learning a common phrase or irregular grammatical structure perfectly well, only to start making errors later on. This could be a sign of progress, and not actual failure.
For example someone learning English might learn the phrase "I caught the bus" early in their study. Later, they would learn that the past tense is commonly formed by adding -ed to a verb, causing them to say "I *catched the bus." Also, after learning the more common meaning of "catch", the student might have trouble associating it with getting to a bus on time, and cause them to use a completely different and possibly incorrect verb, such as "found" or "reached". The student is making more errors, but is relying on much more complex knowledge and less rote memorization. This is a sign of progress, and not of regression in knowledge.
These differences in learning language compared to learning other subjects makes the idea of software that would adjust the course programmatically difficult, and most likely entirely impossible. Of course, I'm still convinced that there are better uses for software in language learning than just as a means of presenting tutorials and providing communication. There are a number of companies already working in this area.
Prerequisites
Anyone who went through high school understands prerequisites. To take calculus, you first have to pass trigonometry. To study Chemistry 102 you first have to pass Chemistry 101 or an equivalency exam.
The model can even be applied at a smaller level. Before studying combinatorial methods, you first need to demonstrate an understanding of set theory. The order of prerequisites suggests a fairly obvious order for the curriculum.
eLearning
This kind of analysis is used in some e-learning software. A colleague working at one such company described how their software tracks a grade school student's performance in various subjects. If the student is struggling in a particular subject, the software predicts what future subjects the student will have trouble in, and where the student can most effectively improve. This could help prioritize efforts to help that student. So for example, the fact that Johnny has difficulty spelling the word "thorough" might be less significant than his problems with memorizing multiples of twelve.
At first glance, this would seem to be applicable to language learning. I certainly was interested in the idea, given my background in workflow and business process analysis. When my colleague described his company's software, I felt I must be on to something in thinking it could be applied to language. As the student learned new grammar and vocabulary, the software could monitor progress and review those subjects that were lacking. For example, if a student of English was having difficulty with irregular past tense (such as "catch/caught"), the system could review the regular past and highlight the irregular exceptions. Unfortunately, there's one problem with this idea. Language learning doesn't work that way.
Acquisition vs. Learning
First off, the way that people learn a language is significantly different from how we learn other things. It would not be unusual for someone living in a foreign country to pick up at least some skill in the local language just through exposure to it. By comparison, it would be extraordinary for someone to gain insights in probability just by hanging out with mathematicians on their days off, but never attending classes.
Secondly, measuring progress in language is not necessarily the same as measuring progress in a subject like math. It is not uncommon for a language student to start learning a common phrase or irregular grammatical structure perfectly well, only to start making errors later on. This could be a sign of progress, and not actual failure.
For example someone learning English might learn the phrase "I caught the bus" early in their study. Later, they would learn that the past tense is commonly formed by adding -ed to a verb, causing them to say "I *catched the bus." Also, after learning the more common meaning of "catch", the student might have trouble associating it with getting to a bus on time, and cause them to use a completely different and possibly incorrect verb, such as "found" or "reached". The student is making more errors, but is relying on much more complex knowledge and less rote memorization. This is a sign of progress, and not of regression in knowledge.
These differences in learning language compared to learning other subjects makes the idea of software that would adjust the course programmatically difficult, and most likely entirely impossible. Of course, I'm still convinced that there are better uses for software in language learning than just as a means of presenting tutorials and providing communication. There are a number of companies already working in this area.
Friday, October 16, 2009
Learning about Learning about Language
This is not an expert analysis of the subject of using technology to teach foreign languages. Rather, it consists of my notes as an experienced software professional who has studied a number of foreign languages and has a passing interest in linguistics. In trying to make sense of the subject, I have started reading anything I can on second language acquisition in general.
My interest is based on little more than a feeling that the enormous advances in communications in the computer world have not been applied extensively to second language acquisition.
I started with a few vague ideas of what could be accomplished.
I now already realize that the first observation is very wrong.
My interest is based on little more than a feeling that the enormous advances in communications in the computer world have not been applied extensively to second language acquisition.
I started with a few vague ideas of what could be accomplished.
- Software that can manage a curriculum in language study in response to the student's progress.
Various e-learning systems track students' progress in order to predict where students may run into problems in the future. Would it be possible to find errors early in language learning, and determine what concepts need review, or even to direct the student to the next useful lesson, based upon existing skills and the objectives in the target language? - Software that can present target language content that is likely to be comprehensible, instructive, and of interest.
So much language instruction software provides its own proprietary content, consisting chiefly of grammatically correct phrases, but not necessarily authentic examples of phrases likely to be encountered in real life. Since the internet provides a huge source of ready-made content, is it possible to search for authentic content, providing a much greater variety of examples of the language for the student?
I now already realize that the first observation is very wrong.
Subscribe to:
Posts (Atom)
