Can Computers Replace Journalists?

New software developed by computer science and journalism professors at Northwestern University, and licensed and distributed by Narrative Science of Evanston, Ill., may be bad news for some writers. The software can generate news stories from hard data inputs. “There’s no human author and no human editing,” says Stuart Frankel, Narrative Science’s CEO. “But the stories sound really good.” So far, the software has been applied primarily to sports statistics, but Frankel sees applications in medicine, crime and finance as well. (Note to self: consider a new career, soon.) [%comments]


TheChacoTaco

This is an interesting concept, however, it would be near impossible to convince the market as a whole that a computer program would be a better journalist than a human being. The general human pride would prevent many companies from jumping on this bandwagon, however, when some company is grasped by desparation, they will probably start this revolution.

Brett

Well somebody will have to input those facts, right? In sports it would be easy enough with a list of standard stats and some highlights quickly added in, but I can't imagine a program-written article will be all that informative when it comes to crime (how many different perspectives/minute details come into play) and other non-statistically based stories (i.e. infinite variables).

One thing that I would look forward to with this tech (if it were possible - and I'm sure it's not) would be unbiased reporting. That would be incredible.

lei

More and more, the superficial nature of words is uncovered.

Ian Kemmish

Yesterday provided a nice demonstration of "intelligent" software's ability to amplify small (or even not so small) events. What would happen if the same thing as happened on Wall St yesterday were to happen to the reporting of, say, an exceptionally noisy firework display near the Dagestan border? Or near the Green Line in Cyprus?

TWL

This looks like another variation on mad-libs; template-based stories. With a large variety of already-written stories and story segments your program can pick the likeliest template (come from behind to win) and fill in the numbers. Surely you've noticed how similar most sports "stories" are.

Don't worry about changing careers

@Brett: Unbiased? Unlikely! GIGO applies and the templates will be pre-spun to reflect whatever bias the purchaser requests.

Jeff B

Although it may be hard for us to wrap our minds around it now, I think that the possibilities in this area are limitless; this may just be the first big breakthrough.

I remember a time when we thought there was no way that music could become more portable than a CD player, and when it was impossible to conceive a 100 gigabyte hard drive, let alone a two terabyte drive. Just because we can't comprehend the 'how' doesn't mean that it's impossible.

PS - Brett, I agree with you, it would be amazing if they could get unbiased reporting out of this, that would benefit the industry more than anything.

Jeff B

Take a look at this article, it has more details:

http://www.businessweek.com/magazine/content/10_19/b4177037188386.htm

It has three paragraphs, two of which were written by real people and one of which was written by the software. Try guessing which one is computer generated, I bet you can't.

Wilma

Yeah, and computers will also replace readers.

humberto

who puts the data? humans and is sad if a maching can write better that us.

DS

Right. They thought spell checkers would replace proofreaders, too, and I'm still working that trade, because there's still nothing in any computer program that matches human judgment. I think reporters are safe from computer programs for at least another generation. Not from the internet turning the profession into a low-paid sinkhole, but from computer journalists, yes.

Eileen M. Wyatt

If this is to replace journalists, I assume we're sending Roombas out to gather the data.

Software could probably do the job for the typical Wall Street Journal filler article ("Market [soars/stumbles] on [insert one world event, one financial event, and one random item, regardless of logic].") But how does it handle complex explanations?

So this raises a question: is the purpose of journalism to convey superficial who-what-why-when-where for a slate of events agreed to be "important" or is it to investigate and explain? (Or, to avoid a false dichotomy, where on a continuum between those poles is journalism valuable?)

Steve Neubauer

@humberto

http://en.wiktionary.org/wiki/irony

Eric M. Jones

The truth is that anything that can be described in an algorithm can be done by a computer....and these algorithms can be so complex that they surpass any single human's understanding.

Not that many years ago engineers struggled to imagine how exactly one could use a microprocessor in, for example, an automobile. Now some cars have as many as 100 microprocessors.

Recently I saw a fake cigarette that used a microprocessor to detect when the "smoker" inhaled and simultaneously injected a controlled amount of vaporized nicotine and illuminated the "ash" of the cigarette in a realistically looking fashion. This could be used to wean you off tobacco. Yikes!

There are microprocessors in bullets, birthday cards, pens, underwear...everywhere. And one might have guessed that Mathematical Calculations would be done on computers, but who could have predicted that they would take over writing, photography, cinematography, music reproduction, astronomy, encyclopedias, wristwatches, telecommunications (cell phones and Internet) TV's, navigation (GPS--even in cars and on foot), medical imaging, microscopy, CAT scans, financial transactions, GAMBLING!, plagiarism detection, library searches, Travel booking, PORNOGRAPHY!, numerical machine control, dating and mate selection, auctions and weather forecasting, video games, robotics, mechanical drawing and design, oh yes...and constantly improving microprocessor chip design.

I'm sorry...what was the question?

Read more...

Ray P

I think it would interesting. Think of computer generated stories based on logic sets that reason over very large or potentially abstract data sets that no one is covering anyway. Most journalists want a quick sensational story but the machine will cover more mundane, less sensational, fact or statistically based stories. Just the facts journalism with no slant... what an idea! Where can I sign up for that feed?

John

I can't remember the last time that I read an article written by a true journalist anyway.

Journalism is dead, bring on the computers!! (They can't do any worse.)

- Silver

"The software can generate news stories from hard data inputs."

All they need to do now is overcome the "hard date inputs" problem and the software generated news stories will be indistinguishable from the human generated news stories.

Ghost

But we thought that journalists had already been replaced by "bloggers" and "twitterers," no?

Are there really even journalists now, anyway? It seems that one outlet actually writes a story, and all the others just re-write or re-post the original article, thereby making it only SEEM as if all of the journalists were actually doing their jobs.

avirr

Life is not chess! There have always been sports and business statistics "stories" -- this seems like just another way to generate them.

But most stories are an attempt to synthesize the information and give it useful context. Even reporting on product updates, which I used to do for tech publications. Most of the press releases failed to include vital data, so I had to chase down pricing and compatibility information.

I'm not saying 'never', what I'm saying is that I want to get my news from a human who is doing their best to understand and share the meaning of the event.

SkepMod

Considering the quality of some of today's journalists, this might be pareto improvement for all concerned - readers, computers and said journalists, who will now be free to pursue more appropriate careers.

Gary

If you have ever read a story about which you know a lot or have been interviewed and quoted, you immediately realize that journalists get only about half of the facts right. On top of that they leave out crucial information (although that may be from poor editing), make unwarranted assumptions, fail to recognize their own biases, and often use non-neutral words to describe situations (a mild criticism becomes a "slam").

I can't see how a computer generated story would be worse.