Comments on: The Chinese Room

By: Strategy Showcase: I, We, You | Mister, is this right?

Strategy Showcase: I, We, You | Mister, is this right? — Fri, 15 Aug 2014 20:13:16 +0000

[…] this mathematical thinking reflects a surface level understanding if any at all. Just like the Chinese room thought experimentÂ (see also), we cannot say they understand mathematics just because they can replicate a procedure. […]

By: Math Example – The Chinese Room | morrisonEDU

Math Example – The Chinese Room | morrisonEDU — Thu, 14 Aug 2014 21:57:55 +0000

[…] Meyer recently posted an article entitled The Chinese Room. The moral of the story is: if I give you something written in Chinese and a manual for writing […]

By: Dan L

Dan L — Tue, 05 Aug 2014 20:39:22 +0000

Yay, the Chinese Room on an educational blog! As some of the commenters above though, I am not sure what purpose does it serve here. In the following I wishfully think that everyone knows Chinese Room and Turing test (the idea, that if something *behaves* intelligently, we might as well consider it *being* intelligent).
In the classical setting, CR says: well, something may pass the test, yet still not *really understand* what is going on (so there is no *true thinking* and the test fails its purpose). Intuitively, this is true. But what we still lack is, ahem, understanding of what understanding really is. Ironically, this is why the T. test was designed as it was: to evade fuzzy notions, such as thinking (as in *really* thinking), it goes with “looks like thinking” instead.
What do we do, as teachers? Ultimately, we train our students to respond to challenges in some expected way. Yes, we have some idealized models of mind and such, but after all, we want to see results in student performance. Of course, I am not talking simple silly rote test questions, which only show shallow application of basic algorithms. We want to see true understanding. We do not want a CR operator. Ok. So… how does a student, who really understands the matter, respond to our challenges? Perhaps we need far better challenges etc., but we are ultimately still stuck within the frame of challenge/response interaction and evaluation. We are not allowed to open up our students’ skulls and rewire or test the brains for true understanding directly.
So where are we? Back in the CR. It is in principle (e.g. with sufficient number of sample challenges, even if they are sophisticated) possible to train the correct responses. And it is even possible to still evade proper understanding. The Chinese Room does not say much more than that we need to define well what understanding really is, and that the risk of misunderstanding while responding plausibly is always present. Our challenges simply have to be good enough so that this risk does not matter anymore.

My personal takeaway for the CR in context of AI: It is not the person who is intelligent, it is the whole system, namely the translation rules. Some sort of degree of that intelligence can be devised of the rules size. The more the rules, the less system in the language do they see, the stupider the room is (in terms of “internal” intelligence, as it performs correctly anyway).
This idea of “size of rules-system” also gives me guide to avoid the CR scenario in my classroom. I may consider the student who memorizes loads of special cases as stupider, than the one who works with a general rule. But I can hardly convince him about the shortcomings of his approach as long as they both perform equally well (and some indeed prefer to use memory over thinking). Investing the extra intellectual effort into learning some deeper rules, connections, metaphors, motivations etc., what can be seen as understanding, must be worth it. If it is not, then I can hardly blame the students to look for more efficient ways to get through. There must be a problem, which a student with understanding solves, and the other does not. What else would be the point of understanding? (Enjoying the innate beauty of the subject is important, but not everyone appreciates it)

So, yes. CR is a pretty perfect parable, it illustrates well some fundamental features of Turing’s argument. Basic philosophy of AI in general has some relevant implications for teaching, since both fields have a lot to do with “faking it”. Those who do not think it through are often misled with their intuitive conceptions of what *true* thinking, understanding etc. actually is and what it is not.

By: Joel Patterson

Joel Patterson — Tue, 05 Aug 2014 00:42:21 +0000

I like this parable. Before you condense it, and give it to all your parents, consider whether those parents have science/engineering backgrounds. It’s a pretty complicated picture to envision. I think S/E people would grasp it (if they haven’t heard of it already) and would get your point. But if the parents have less of an S/E background, the complicated parable is likely to bore them and not convey your point.
Have an explanation at hand that is more like the guitar players who can improvise, not just repeat the 5 songs they’ve memorized. Or cooks who can put together a soup without the recipe because they know which spices and foods have good flavors together.
A good explanation is one that a broad audience can relate to. If we’re going to succeed at this we’re going to have to relate to people who have spent years of their lives thinking they don’t get math or science.

By: Dan Meyer

Dan Meyer — Mon, 04 Aug 2014 20:42:33 +0000

Click on through. I found Searle’s original formulation less useful than Greeno’s paraphrase for the purpose of math teachers.

By: Haiwen

Haiwen — Mon, 04 Aug 2014 20:23:44 +0000

Did Greeno also not cite Searle? Searle is alive and well!

By: Howard Phillips

Howard Phillips — Sun, 03 Aug 2014 22:45:21 +0000

I love this parable !!
People whose job is to mend televisions (old style, analogue, with tubes) do the job based on a set of rules or instructions, but are very unlikely to know how a television works. I am aware that “math is not like that”, but understanding is a lot more than understanding how standard algorithms work. It has a lot to do with understanding what all this stuff is for, and without this depth of experience the “have a go” person is not going to get very near any sort of solution. We are wasting time teaching techniques for exam purposes, but we are running the risk of wasting time insisting on students giving explanations for everything. Math education will only improve when the development of problem solving skills is seen as the justification, with the tools being introduced as and when appropriate.
I remember being the man in the room when I was a student studying some aspects of abstract algebra.

By: Sadler

Sadler — Sun, 03 Aug 2014 15:20:38 +0000

Not so perfect in my understanding. Let’s compare the maths student to the symbol manipulator.

It is assumed in this parable that the symbol manipulator has zero knowledge of Chinese, however he/she has some other language, has eyes (presumably) and some rules written in their own language. Normally, when given a range of evidence and a way of experimenting, one will hypothesize and test. Even if the manipulator cannot witness the success or failure of their experiments, they can still test them against the rule book and their native tongue.

You’ve got to remember that Searle intended this parable to refer to computers with zero semantic context which just isn’t possible with people and I would trust anyone in such a situation to come up with their own names for characters (vocab), extend the rules (syntax+grammar?) and apply their own meaning. One could argue that the meaning might be wonky but how by how much?

The maths student, on the other hand, goes into the Chinese room with plenty of context and plenty of fundamental Chinese – if the teacher teaches the rule book without correcting misunderstandings then they’re just a bad teacher.

By: Harry O'Malley

Harry O'Malley — Sun, 03 Aug 2014 15:09:08 +0000

This parable has so many contextual elements surrounding it, it is impossible to know what you mean when you say it is perfect. Perfectly achieving what?
Anyway, let’s use it as a chance to riff. I first encountered this parable 4 years ago reading Steven Pinker’s “How the Mind Works” (which is a stunningly rich read). There he tells us (as does Wikipedia), that the parable was originally written by John Searle to convince others that, although a computer may be able to simulate intelligence and understanding, this would remain a simulation. The computer would not actually understand.
In the book, Pinker challenges this thought experiment from a number of angles.
Firstly, the person in the room can be thought of as our brain instead of a computer. Just as it is easy for us to imagine a computer program not understanding the big-picture point of what it is doing, it is also easy to imagine our brains not knowing the big-picture point of what we are doing either. In this view, our brains simply carry out the tasks of taking in the sounds of verbal language, processing them in a myriad of ways that we don’t understand, packaging them into neurological deliverables that coast toward the action muscles in our mouth and cause us to speak a response. Adding an ironic twist, we do not understand almost anything of how our own brains do what they do. So who really understands Chinese, us or our completely mechanical brains?
Secondly, Pinker calls our attention to the fact that the speed with which the Chinese translation task is carried out may affect our perception of the translator’s understanding of Chinese. The parable works in part because, as we play it through in our heads, we imagine a person tediously and slowly looking at the characters that are input, flipping through pages of a hardbound book (how slow!) looking for the right translation rules, carefully copying the rules from book to page (probably with a quill and ink) in a step by step fashion and finally slipping the response back under the door. This slow process lubricates the path toward concluding that the person doesn’t understand. Now speed this up to the point where the person could answer people back at a fluent conversational speed or better. Does the person (or the room itself, maybe) understand Chinese now? It is not as easy to say no.
For more food for thought on this, the link below provides an example where a computer, namely IBM’s Watson, is being used to perform a cognitive task that we often associate with higher order understanding: creativity. And not just any creativity, but creativity that results in valuable human solutions to specific problems.

https://www.youtube.com/watch?v=EWF4sC5wuqs (follow the link in the video to find out more. fascinating.)

I have a deep goal to teach students to understand mathematics. But a fundamental challenge has always been to figure out what it means to understand mathematics. It is easy to read this parable and think “Ahhh, yes. Someone else understands what I mean when I say that simply becoming fluent in the rules of mathematics is not the same as understanding.” It seems to perfectly assimilate into our current views (I suspect that’s what you meant by perfect, Dan.) But Pinker’s arguments helped me turn this into a deep exercise in accommodation, challenging my beliefs. It is possible (probable, even) that a large part of understanding is the accumulation of tons and tons of specific examples of different mathematical phenomena over and over again.

By: Eileen D

Eileen D — Sun, 03 Aug 2014 14:29:45 +0000

Always my greatest fear as a math teacher! I want to teach concepts not tricks. Each year as we close in on end of the year testing it is so hard not to teach them how to “roll over.”