Whether or not the title of this week's post is funny is debatable, but at least it's based upon a principle of humor that could lead a human reader to an amusing and humorous interpretation. You see, robots are machines, and since machines are manufactured by having components held together by screws, telling a robot to “go screw yourself” (especially a robot who has been put in charge of converting humans into slaves for the benefit of the system that built the robot) lends itself to the jocular ambiguity of a pun by referencing a traditional epithet that has long been applied to reprehensible leaders who would exploit their lackeys (and that epithet being, of course, that if the leader is determined “to screw” his minions, he should rather emphatically go satisfy his lascivious requirements through an exclusive and solitary onanism).
Now here's an important question: Would a robot find the preceding paragraph funny? If you are human, you might respond to this question by saying the question itself is absurd. Because robots are incapable of experiencing human emotions and laughter is an emotional response triggered by a comic awareness unique to human sensibility, the question as to whether a machine can determine 'if something is funny or not' is meaningless. This is to say that humor is subjective; in order to decide if anything is funny, the circumstances require a human subject capable of personal and intuitive response. While people can argue over whether or not any specific joke is funny, we are likely to find near complete unanimity if we are arguing instead whether or not machines are capable of appreciating humor.
Now a thought experiment: Suppose an eccentric billionaire came to you and offered you an insane amount of money to create a robot that would laugh at his jokes. This leads to an ethical dilemma, right? Do you create a robot that seems to laugh at the billionaire's jokes in order to become fabulously wealthy, or do you lose your chance at having all that wealth by honestly admitting to the billionaire that because machines cannot really laugh, taking his money for a laughing machine is inevitably an act of fraud?
Some people might respond to this by saying, “If someone has more money than brains, then you should go ahead and take the money and build a robot that plays a recording of laughter when the billionaire uses a particular tone of voice that indicates he is being sarcastic. It doesn't matter if the robot has no more sense of what the billionaire is saying than any typical department store mannequin, if someone is willing to pay billions of dollars for a laughing robot, it doesn't matter if the robot can actually laugh at what is being said, the billionaire only needs to believe it is laughing at his jokes. If a robot played a laugh track at intervals that gave the impression it was laughing at the appropriate times, it would be up to the billionaire to decide if he was being ripped off.”
Other people would respond by saying, “Taking money for one thing and delivering a product that does something else is fraud. It doesn't matter how much money is involved. Because robots cannot interpret humor, a uniquely human ability, taking money for “an authentic laughing robot” would be dishonest no matter how satisfied the billionaire would be with the results of a machine that would produce laughter at convincing intervals.”
Suppose in your conversation with the billionaire, you ask him, “Why do you want a laughing robot? Why not just hire a panel of clowns to laugh at your jokes? Surely you can afford to hire people to laugh at what you say.” And the billionaire responds, “I don't want people to laugh at what I'm saying merely because I'm paying them to laugh at what I'm saying. I have tried this method in the past, and when I pay people to laugh at what I'm saying, they laugh at everything whether it's really funny or not. What I need is an objective measure of humor. I know that not everything I say is funny, but some of it is. I want a device that can tell me when I've said something funny. If I had a robot, it would be completely objective in determining whether something is funny or not because robots cannot be bribed.”
This, then, is crux of the problem: The human desire for “an objective measure” of “something that cannot be measured objectively” cannot take precedence over the logic of measurement. That is to say, a fundamental principle of rationality maintains that “a motive to have specific information” cannot supersede “the reasoning that can provide the information.” To be blunt, I'll put it this way: Anyone who claims they can objectively measure anything that depends upon human subjectivity is either a fraud or fool.
In the example above, the billionaire's access to endless wealth is irrelevant to obtaining the information he wants. Because humor is by definition “subjective,” an “objective” measure of humor cannot be had at any price. Money can buy a lot of things; it can purchase agreement, but it cannot shop for authenticity for things that cannot be authenticated. Numbers are very good at describing things that can be measured – the distance between New York and L.A., the weight of an elephant, and average salary of postal workers in the United States. What numbers cannot tell us is qualitative information than in simply not reducible to quantification. The distance from New York to Los Angles is 2,778 miles; but who can save authentically that “It's too far to walk”? The subjective determination that “It's too far to walk” is based on a wide variety of human motives and conditions. It wouldn't be too far to walk if you had the right motivation to walk it. Even if 99.9% of a survey of the general American population declared that “It's too far to walk,” you might walk the distance if you had the right motivation. We might use statistical data to learn that the average salary of a postal worker is $48,380. Whether you believe they don't make enough money, make too much money, or make the right amount money, your belief can be justified by a wide variety of arguments, but your opinion can not be quantified into the “one right answer” because there is no one right answer. Opinions are subjective. Only numbers are objective, and numbers can't have opinions.
To recap: Objective data can provide information that can be authenticated. Opinions can be informed by objective data, but because interpretation is a subjective human response, opinions can only be justified, they cannot never be authenticated. Anyone who claims that they can provide an objective answer to a question that requires a subjective response is being disingenous. If vasts amounts of money are involved, it is my informed opinion that fraud on a massive scale is inevitable.
If you are a human, you may interpret this essay as an inditement of the corporations that are currently conspiring “to grade” student writing with “intelligent software.” I would argue that while the software may be intelligent, the people who believe in the results are not. Whether those who are either developing the software or buying the software are being entirely honest about their motives for saying they believe in the results depends entirely on how much money they are being given to authenticate the results.
Was this essay “well-written”? I don't suppose any machine could offer an opinion about this one way or other. Software can offer information that can make predictions on the quality of writing based upon the metrics of sentence length, vocabulary usage, and grammatical conventions; Software cannot “read” writing for content and recognize subordination of ideas, rhetorical fallacy, or metaphorical language. I'll believe in software that can make accurate predictions of the greed of corporate hucksters, short-sighted politicians, and budget-conscious school administrators long before I'll ever be able to accept the existence of software that can judge “good writing.”
Keep thinking rhetorically, and I will return next week.