An interesting short conversation has just been posted on Philosophy Bites with Daniel Dennett, the American cognitive philosopher, in which he discusses a thought experiment invented by another philosopher, John Searle, called the Chinese Room (you can listen to the conversation here).
Searle devised the experiment over thirty years ago to argue that no computer, however sophisticated, could ever be said properly to understand a natural language.
The experiment goes as follows: Imagine a computer so advanced that it could pass the Turing test and convince an interlocutor that it was human. This particular computer can hold a conversation with a Chinese interlocutor in another room. The Chinese speaker types in a string of questions or remarks, and from the replies is persuaded that he or she is talking to another human being.
Then imagine that a human, an English-speaker with no knowledge of Chinese, replaces the computer. This human, however, has access to the computer manual (written in English) and the code, which enables him or her to perform the same input-output operations as the computer. In this way, the English-speaker can interact with Chinese speakers outside the room in the same way as the computer. But the human cannot be said to understand Chinese.
Neither, argues Searle, can the computer. He maintains that there is no work of understanding being done at all on the part of the computer. There is only a formal (syntactic) manipulation of symbols, with no regard anywhere in the system for their semantic content – in other words, what they actually mean.
English teachers learn Searle’s lesson very early – under certain circumstances a student can provide answers that give the appearance of understanding when there is in fact a fundamental or radical misunderstanding; they are following the logic of the exercise, or of sub-routines of their own invention, not processing the language.
Last week I was teaching a group a little gerund and infinitive; in exercises and free production students got some wrong, some right, as might be expected. It is a thorny area. But on Friday, in a review of the week, it emerged that one student was working to a perfectly reasonable but as it happens entirely incorrect hypothesis (regarding whether an activity was indicated or a state, if I understood) which, given that he remembered a few anyway, happened to be providing the correct answer about as often as you’d expect.
Similarly with vocabulary – an exercise can provide a syntactic framework within which symbols can be manipulated without any actual understanding on the part of the student. And, odd though it may seem, understanding is not always a student’s primary goal; sometimes they only want conform to the group, not to hold things up, not to appear more stupid than their peers.
All this assumes anyway that students should be parsing questions and answers for explicit grammatical structures in order to function in the language. We all know from rudimentary experience that this is not the case. You can interact very happily with an extremely low grammatical competence, but with a good awareness of contextual clues – body language, visual prompts, and so on. Conversely, you can give the impression of having understood (also in your mother tongue) when you have not in fact been listening to a word.
So we continue to pass messages in and out of our various boxes, and hope, if we think about it at all, that the processing going on in any given room is what we think it is. Because we only have a few rudimentary outputs – known collectively as language – to work with.