The Chinese Room
There is a thought experiment called The Chinese Room used to determine if machines like computers have an understanding of the tasks they perform. The experiment runs like this: Suppose a computer is constructed and it behaves as if it understands the Chinese language. It takes written Chinese questions as input and, following its programming instructions, produces Chinese answers as output, as if the asker is interacting with a Chinese-speaking person. Now suppose inside this computer is an English-speaking human, with no knowledge of Chinese and a “program” of instructions in English. His program manual is designed to allow him to look up the input Chinese characters, process them according to a set of English rules, and produce the appropriate Chinese characters as output. The output is the same whether the computer is electronic or human-based – each follows a set of instructions to produce the desired output. Because the human can produce the desired output without understanding a word of Chinese, the electronic version cannot be said to “understand” Chinese either.
Yesterday, I applied the thought experiment in the real world.
My first test was during lunch at my favorite sandwich shop. I ordered a turkey on wheat with mayo, tomato, and lettuce. As I placed the order, I thought about the “machine's” input – my order – and the output – my sandwich. A few moments later, the preparer handed me my order, I paid, and I sat at a table to eat. Treating the transaction as a black box, I thought, “Could my sandwich have been produced by a computerized machine following a set of sandwich preparatory instructions with the same results?” The sandwich barista did not need to “understand” sandwiches or culinary subtleties, only follow a few simple sandwich assembly rules. Nor did the cashier need to “understand” the sales transaction, she only needed to push picture buttons representing items I ordered and allow the register coin dispenser to dole out change based on the cash I gave. Conclusion: the whole process could be thought of as a type of Chinese Room where no understanding of the inputs is required to produce the relevant outputs.
I next tested the experiment at my work's weekly status meeting. The meeting is attended by a dozen people; workers each give the project manager a couple minutes of status on the past week's tasks, while the project manager types a summary into his tracking spreadsheet. The meeting is concluded with the project manager encouraging attendees to keep focused and work hard to meet his schedules. Again treating the situation as a black box, I thought, “Could this meeting have been run by a electro-mechanized entity with the same results?” The inputs are statuses from meeting attendees, the outputs are a rote compression of those statements into a computerized database, followed by a recitation of standard workplace expectations. Yes, the meeting is classifiable as a Chinese Room with no thought or understanding required.
I looked around my workplace to observe typical tasks. Jim enters sales figures into a spreadsheet, adding, subtracting, and averaging numbers. The spreadsheet does all the work, all Jim does is transfer numbers from one medium to another and apply a few rules for calculations. He could easily be replaced by a non-thinking scanner and computer applying the same rules. Sandy schedules meetings given lists of attendees; she adds the attendees based on availability as tracked in the scheduling tool. Again, substitutable by a non-thinking computational entity following an availability algorithm. Andy calls customers inquiring if they need refills on supplies; a job which could be handled by an email and website interface. All Don does is sit in his cubicle and surf the web; he produces the exact same output as a machine that is switched off.
On the drive home, I observed driving is nothing but routine decision making based on sensory input; decisions could be compiled into a short list of program instructions – accelerate, turn, slow, stop – based on optical and sound stimuli. Dinner preparation is reducible to following a recipe without needing to understand anything about flavors, consistencies, or their combinations. Even when watching a favorite sitcom I don’t need to comprehend what is or isn’t funny, a laugh track informs me. No thought required.
You might have noticed I ran the thought experiment in reverse – essentially determining if a process “needs to” understand versus determining it “can” understand. But that is my point: there are so many aspects of our lives that have become routine, requiring no thought, they can be viewed as a type of Chinese Room, completely replaceable by a non-thinking machine. Operating without needing to understand has become normal human procedure.
And we wonder why people do not think critically, cannot sort out phony information from facts, and fail to cast ballots that are in their best interests.