Fallacies in Searle's Chinese Room Argument

The question of whether or not a machine is capable of intelligent, human-like thought is one that has plagued philosophers and scientists for centuries. Descartes considered the question in 1673 and determined that the sign of intelligence is the ability to understand language, and indeed today, a machine's ability to understand language is the standard against which we test its capacity for thought within a machine. Father of modern computer science, Alan Turing, formalized this standard, creating what is now referred to as the Turing Test. The Turing Test involves two humans and a computer. One human questions the other human and the computer, and attempts to distinguish which responses belong to the computer and which belong to the human. If the interrogator is unable to accurately distinguish the computer from the human, then the computer is said to pass the Turing Test. Many people believe that a computer could be built that is capable of passing the Turing Test and thus exhibits thought just as the human mind does. These people are proponents of strong artificial intelligence.

American philosopher John Searle attempted, in his 1980 article "Minds, Brains, and Programs," to prove that a computer running a program does not enable the computer to understand language, and thus no real intentionality is being exhibited through the computer's ability to manipulate language effectively. At the core of his argument is the thought experiment, known as the Chinese Room. Searle asks us to imagine a computer that behaves as if it understands Chinese, with the ability to take Chinese characters as input and, following a set of rules, output appropriate Chinese characters. This computer performs its task so convincingly that it easily passes the Turing Test, and so supporters of strong artificial intelligence conclude that the computer understands Chinese just as a fluent speaker does.

Now Searle asks us to imagine a person sitting in a room with a rulebook of Chinese characters. A fluent Chinese speaker sitting outside of the room enters input and passes it through into the room, where the person inside the room, though not knowing any Chinese language at all, consults his rulebook and outputs the appropriate text. Searle claims that the person inside the Chinese room does not understand any of the text he is receiving or outputting, but would be able to convince the Chinese speaker outside of the room that he, or the Chinese Room, does, as his responses make perfect sense due to his detailed rulebook that says "when you see a squiggle like this, write a squiggle like this." Searle says that because of this, the Turing Test is not sound, as the man in the Chinese room convinces the Chinese speaker that he understands the text when he is actually simply a mindless manipulator of symbols. For this reason, Searle claims that simple grammatical and syntactical manipulation of language does not necessitate or implicate the existence of a mind or intentionality, as there is no meaning behind the language being processed. A mind requires semantic abilities to interpret symbols for meaning.

One of the more common rebuttals to Searle's claims is called the Systems Reply. The Systems Reply states that Searle's claims are misleading, as they ask us to focus on the person sitting inside the room. Searle claims that this person does not understand Chinese, which is a fair claim and more than likely and accurate claim. The problem, however, arises when you examine the system as a whole. When you put together the rulebook and the symbol manipulator (the person inside the Chinese room), you form a system that does understand what it is doing semantically. The symbol manipulator is merely a part of the larger system. Searle, however, responds to this point by stating that even if the person inside the room were to commit the rulebook to memory, he would still be mindlessly manipulating symbols: he is only taking the symbols he sees before him, matching them with the ones stored inside his head, and outputting the appropriate symbols in turn. Searle claims that now, the manipulator understands no Chinese, just as the system does not as well, since there is nothing more to the system than the manipulator.

The best response to Searle's defense is an extension to the Systems Reply popularized by Daniel Dennett in his Consciousness Explained. Dennett claims that Searle's entire argument is misleading, as it all relies on the fact that one could make a machine so simplistic that could so easily pass the Turing Test. It is highly unlikely that a system consisting of only a man and a look-up table for a rulebook would be able to have a conversation in Chinese that would easily convince a fluent speaker that there was another fluent speaker inside the room outputting the conversation. Searle's whole argument is predicated on this assumption, and it is very likely that this assumption might be false under all conditions.

If we extend the system such that it contained all the detection systems required to ensure perfect, sensible responses with each output, the idea of the system passing the Turing Test quite so easily becomes much more insecure. Indeed, for the Chinese Room to easily pass the Turing Test in the way it is described by Searle, it would necessitate either many, many, mindless manipulators inside the room, or a tremendous amount of time for the manipulator to produce an answer to even the most basic questions. Both of these conditions are Chinese Rooms that would pass the Turing Test, but they are irrelevant, as both systems would be so complex that it would not be obvious that no clear consciousness was present, as the system would entail that concepts be weighed against one another and possible answers be formulated while the system slowly analyzes the semantics involved with the symbols as it prepares its output. If the system performed these tasks faster, then it would appear that the system perfectly resembled a Chinese speaker.

The "What More Do You Want?" objection often brought up by supporters of strong artificial intelligence enhances the Systems Reply. Essentially, scientists say that if they were able to build a machine like those in the science fiction movies, or perhaps one like the one described above in Dennett's extension to the Systems Reply, then what more could Searle possibly be looking for? If the machine could pass any test that humans could devise, then any discussion of differences between it and a human mind are of no interest to science. Searle argues that what separates mind and machine is the existence of consciousness and semantic intentionality, but Dennett is able to provide a machine that is capable of exhibiting both of these features.

The fallacies that exist within Searle's Chinese Room argument severely cripple it. To assume that a machine like the simplistic one he documents would pass the Turing Test is ludicrous and disingenuous, as is the idea that a system capable of passing the Turing Test would obviously not exhibit consciousness, or at the very least semantic intentionality. Until a stronger argument against the ability for a machine to think and exhibit intentionality and consciousness is formulated, we must assume that it is a scientific possibility.

Sources

Dennett, Daniel C. Consciousness Explained. 1st ed. New York: Library of Congress Cataloging-in-Publication Data, 1991. 435-440.

Rachels, James. Problems From Philosophy. New York: McGraw Hill, 2005. 91-98.

Searle, John R. "Minds, Brains, and Programs." The Behavioral and Brain Sciences 3 (1980). 10 Dec. 2006 .