Siri is just the beginning.

Soon enough — like it or not — we are all going to be talking a lot more with computers, at least according to Brian O’Neill, a professor of computer science at Western New England University. They will be tutoring our students and tending to us in our old age. And a big part of their job will be telling stories.

It’s about more than just entertainment. Teachers use storytelling in all subjects, including science and math classes. Doctors say understanding patients’ stories improves health care. There’s even evidence a good yarn can treat high blood pressure.

But O’Neill says there is a problem: Computers are boring storytellers.

“Right now we can get them to tell things that maybe are like very simple stories, but they’re not interesting,” he says. “And we’re going to want those stories to be interesting.”

The first step toward fixing this problem, as O’Neill sees it, is to teach computers how to recognize one of the key features of an entertaining story: suspense. To do that he needs a simple definition of suspense that he can program into a computer.

Here is the one he likes most: “We feel suspense when we have less and less hope for a character escaping a bad situation. The less hope we have, the more suspense we feel,” he says.

Last year O’Neill wrote a computer program called Dramatis that he hoped, operating from that narrow definition, would be able to detect suspense in a Hollywood blockbuster. Dramatis can’t actually watch a movie, but it can read a dumbed-down version of a script.

When the hero reaches a dangerous point in the plot, Dramatis asks itself: “If I were in this position, what could I do to get out?” In essence, it tries to imagine an escape plan. And the more implausible that plan seems, the more suspense Dramatis registers.

O’Neill and his colleagues put his computer program to the test by feeding it suspenseful movie scenes, including Grace Kelly’s confrontation with a murderous neighbor in 1954’s Rear Window, a very fateful moment for Professor Dumbledore in 2009’s Harry Potter and the Half-Blood Prince and James Bond’s treacherous poker game in 2006’s Casino Royale.

Dramatis began dutifully reading and calculating.

In the Casino Royale scene, for example, Dramatis tracks 007 as he is poisoned by his archenemy during a poker game. Bond excuses himself, then swallows copious amounts of salt water to try to induce vomiting. But this first plan fails.

There is now less hope for a peaceful resolution for our protagonist and, accordingly, Dramatis increased the scene’s suspense rating.

Next, Bond staggers out to his gadget-rigged Aston Martin and phones MI5 in London. The spy agency’s doctors tell him to use a defibrillator to jump-start his heart immediately before he passes out. Bond hooks up the leads, and tries pressing the button to release a charge. But, over and over again, that doesn’t work.

Dramatis, Brian O’Neill says, starts “feeling more suspense because we know what the plan was,” and that plan is proving unsuccessful.

Finally, we learn that a wire is accidentally loose. And before Bond can reconnect it, he collapses from the poison.

“Now we have a much harder task, and we’re going to feel even more suspense now,” O’Neill says, “because it’s much harder to come up with somebody else who can save James Bond.”

In the end, Bond’s sidekick Vesper Lynd zaps him back to life. But as things became worse for Bond, the Dramatis delivered higher and higher suspense ratings.

Since Dramatis had only been reading a written description of the scene, not watching it, O’Neill was later able to tweak the script to create a new version where Bond’s escape is easier. O’Neill also created alternative, easier-to-escape versions of the Rear Window and Harry Potter sequences.

Finally, O’Neill had both Dramatis and human test subjects read both versions of the movie scenes and compare them. (Since Dramatis can only read — not watch — a movie, O’Neill gave the humans written descriptions of the movie scenes to level the playing field.) At the end of the experiment, the computer and the people were in agreement: The versions where heroes escape easier were less suspenseful to both readers.

So, does this mean Dramatis understands suspense as a human does?

Livia Polanyi, a professor of linguistics at Stanford University who specializes in narrative says this work is a meaningful advancement for computerized storytelling; Dramatis does a lot right.

But, she says, the computer still does not fully grasp what makes us feel suspense. It cannot detect music, anguished faces and, most importantly, the way we humans connect with the characters. If we don’t care about them, we don’t care what happens to them.

“I think we have to really account for what actually causes us to identify, to empathize and to care,” Polanyi says.

Teaching empathy to a computer will take considerably more time, but for Brian O’Neill and other researchers trying to develop a less boring computer, Dramatis is a good first step.

Copyright 2015 NPR. To see more, visit http://www.npr.org/.