Real-Time Automated Essay Writing?

When I first tried EssayTyper, for just a moment it chilled my blood. Of course, it’s just a little joke; but I hope students everywhere will be sophisticated enough to see that, because a person who was unusually naive, lazy, and ignorant just might mistake it for a computer program that will enable you to type out custom-designed essays on selected academic topics, even topics you know nothing about, even if you can’t type. The EssayTyper home page presents a box saying:

Oh, no! It’s finals week and I have to finish my American Civil War essay immediately.

You can type in a replacement for “American Civil War”; whatever you please: “praseodymium” or “eagles” or “Cole Porter” or “phonetics” or “Chronicle of Higher Education” or “lingua franca”—anything you could imagine someone being expected to write an essay on.

If then you click on the pencil icon on the right hand side, you get what appears to be a word-processor page with a centered header providing a fashionably absurd postmodernist title for your essay: “The Fluidity of Praseodymium: Gender Norms & Racial Bias in the Study of the Modern ‘Praseodymium,’” or maybe “Truly Eagles? The Modern Eagles: a Normative Critique.”

All you have to do after that is type. Type anything. Rattle your fingers around on the keyboard like a child pretending to type. Have your kitten walk on the keys. Tap the space bar. It doesn’t matter. Text will appear, bit by bit: coherent, sensible text saying true things about your chosen subject. Not very imaginative, but undeniably accurate and probably worthy of a B grade.

Now, we already know that the humor-detection module in our species is not innate, so there is a real chance of my being disappointed in our students: There may be some who think EssayTyper is more than a joke. I continue to hope otherwise, partly because humor sensitivity is generally stronger in the young, and partly because I simply don’t want to live in a world where this tool might be used to create essays that might be turned in for me to grade.

EssayTyper is actually (to give the game away completely) a front end to Wikipedia. When you type your subject in on the underlined part of the initial box, it simply looks those words up using the Wikipedia search function. If there is no Wikipedia page with that title, it warns you that it can’t help. But if there is one, it goes to it and starts blurting out the text of the article, chunk by chunk. The more you rattle the keys, the more it puts on your screen.

EssayTyper is less intriguing than Eliza, an ingenious piece of programming that was originally intended to demonstrate shallow-level simulation of human conversation but ended up unexpectedly demonstrating human gullibility. EssayTyper is a cute little piece of recreational programming fun, but underlying it is nothing more than an automated Wikipedia copier.

So even for students who think they can get away with turning in unmodified Wikipedia articles as term papers, EssayTyper would be an unneeded middleman. Screen-scooping selected text directly from Wikipedia itself would be quicker.

But as I said, when I first saw it working, for a minute or so I was scared. It isn’t real, and it doesn’t pretend to be, but what if it were? What if, five or 10 years from now, sophisticated programming permits generation of highly plausible text on arbitrary subjects that has been skillfully rearranged from its various online sources, with random words replaced sensibly by synonyms, so that plagiarism-detecting algorithms report nothing untoward? What if machines can one day write convincing original term papers that have not gone through even one human brain before being dumped to the printer?

I like to think I would be able to tell. But how can I know? In the 20th century people used to bet that no machine could ever beat a grandmaster at chess, and they look silly now. I’m betting that computer simulation of text on academic subjects cannot improve fast enough to overtake my cheater-detection abilities. But I could lose. Twenty-five years ago I would have bet that high-quality automatic recognition of arbitrary phrases over a phone line would take 50 years to accomplish. Yet I spoke to a speech-recognition algorithm on the phone just this morning, in a call to the British income-tax authorities. Its performance was flawless, and similar systems have been operating for years. I was wrong there, and I could be wrong again.

Return to Top