Every­one thinks they can write. Even, it would seem, robots.

But do these auto­mat­ed writ­ers real­ly have the abil­i­ty to pro­duce con­tent to rival the work pro­duced by those of us who write for a liv­ing? Should we see them as a threat?

Like­wise, whether you are a brand man­ag­er with­in retail, trav­el, finance or any oth­er indus­try, will this affect you and what, if any­thing, do you need to be think­ing about now? Could you actu­al­ly con­sid­er using robots instead of humans to get your con­tent writ­ten?

Well, it’s time to see what they’ve got to offer as, in what is believed to be a world-first for jour­nal­ism, robot-gen­er­at­ed sto­ries have been pro­duced by The Press Asso­ci­a­tion.

The auto­mat­ed press ser­vice set up by PARADAR — is cur­rent­ly tri­alling com­put­er-gen­er­at­ed-data-dri­ven con­tent, fund­ed by a grant from Google’s Dig­i­tal News Ini­tia­tive.

The plan: to cre­ate 30,000 localised sto­ries a month from data using Nat­ur­al Lan­guage Gen­er­a­tion soft­ware.

At the end of Novem­ber 2017 a pilot, involv­ing 35 region­al titles from 14 pub­lish­ing groups includ­ing Archant, Inde­pen­dent News and Media and John­ston Press, result­ed in mul­ti­ple ver­sions of four sto­ries being dis­trib­uted. These have since appeared in week­ly and dai­ly titles both online and in print.

What This Means for The Press Association

In its own words: “Press Asso­ci­a­tion was con­ceived as a Lon­don-based news gath­er­ing ser­vice for the provin­cial papers … we are trust­ed because we are fast, fair and accu­rate. Today, much of the con­tent peo­ple read, see or hear con­tin­ues to orig­i­nate from PA.”

Ulti­mate­ly, it gath­ers data and facts that are then sent out to jour­nal­ists who should source quotes and local­ize this infor­ma­tion, turn­ing the bare bones into a sto­ry suit­able for their indi­vid­ual pub­li­ca­tion. So, in the case of com­put­er-gen­er­at­ed sto­ries, this could offer an effec­tive way of the Press Asso­ci­a­tion send­ing out a lot of infor­ma­tion very quick­ly.

It is impor­tant to note here that reporters are still at the core — mere­ly using the infor­ma­tion they receive as the start­ing point for sto­ries.

So, the real ques­tion is, could a robot ever be trained to write the full piece? And, will they be replac­ing humans at news desks up and down the coun­try?

The short answer is no.

More often than not, we want much more from an arti­cle than just the who, what, where, when and how — it must also take us to the emo­tion that lies beyond the fact. You don’t, for exam­ple, just want to know that a school excelled in its A‑level results this year, you want to hear from the child who beat the odds to get top marks and is now set to attend Cam­bridge Uni­ver­si­ty.

This human ele­ment is key and it is why we can’t under­es­ti­mate the impor­tance of hav­ing a human write the con­tent. There is a cer­tain skill to writ­ing — par­tic­u­lar­ly copy that needs to be enter­tain­ing, engag­ing or per­sua­sive — that goes well beyond typ­ing words on to a page. To put it blunt­ly, sole­ly data dri­ven con­tent is dull. It lacks the emo­tion and con­text that us writ­ers could – and should — inject into a sto­ry.

We Still Need Human Writers

Recent­ly, a new chap­ter was writ­ten for the Har­ry Pot­ter series titled: Har­ry Pot­ter and the Por­trait of what Looked Like a Large Pile of Ash. But this was not fan fic­tion, nor was it writ­ten by J.K Rowl­ing her­self — it was actu­al­ly typed up by a pre­dic­tive key­board.

After feed­ing sev­en books through the com­put­er, lines such as, “He saw Har­ry and imme­di­ate­ly began to eat Hermione’s fam­i­ly”, and “‘Not so hand­some now’, thought Har­ry as he dipped Hermione in hot sauce,” were pro­duced.

The fact that your con­tent needs to make sense goes with­out say­ing … but not when it comes to robots it seems.

This isn’t the only thing to con­sid­er, how­ev­er.

Injecting Humor

If you get it right, you will reap the rewards from humor­ous con­tent. Peo­ple like to be enter­tained — and will share and inter­act with your con­tent if they feel they have been enter­tained. But can a robot be fun­ny?

Dur­ing his time as New Yorker’s car­toon­ist, Bob Mankoff devel­oped an inter­est in the cre­ative poten­tial of arti­fi­cial intel­li­gence. After launch­ing the car­toon cap­tion con­test and receiv­ing up to 10,000 entries a week he attempt­ed – along with Microsoft and Google’s Deep­Mind — to devel­op an algo­rithm that could dis­tin­guish between those that were fun­ny and those that weren’t. How­ev­er, he even­tu­al­ly declared it a ‘dead end’.

Yet he did form Bot­nik stu­dios and cre­at­ed the tool used to cre­ate the Har­ry Pot­ter chap­ter. This tool takes the essence of a pub­li­ca­tion or top­ic — such as David Attenborough’s Blue Plan­et — to cre­ate some­thing ‘com­plete­ly absurd’.
A tool which, although fun to play around with, won’t cre­ate con­tent wor­thy of plac­ing on your web­site in a hur­ry.

Being Aware of the Details

No mat­ter how clever they seem and how much they can already do that you prob­a­bly would nev­er have expect­ed, you can’t train a robot to have news sense. Take the exam­ple of this news sto­ry cre­at­ed by a robot for PA: ‘Most babies are born to mar­ried par­ents in Bournemouth, fig­ures reveal’.

In this data-heavy arti­cle, the robot wouldn’t know if the mater­ni­ty ward was due to be closed, for exam­ple. It’s often extra infor­ma­tion such as this which adds an impor­tant angle to the sto­ry, mak­ing it more news­wor­thy.

Data can orga­nize infor­ma­tion, but jour­nal­ists turn it into a sto­ry.

Recognizing Context and Emotion

The words shouldn’t only tell us the facts, they should bring a sto­ry to life and tap into our emo­tions.

Take the dif­fer­ent approach­es to this sports sto­ry, for exam­ple.

The robot ver­sion begins: Mar­cus Paige scored with nine sec­onds remain­ing in the game to give North Car­oli­na a 72–71 lead over Louisville.

While the human ver­sion, writ­ten for ESPN, opens: Mar­cus Paige ignored the pain in his twice-injured right foot, put his head down and drove toward the rim.

This sto­ry­telling ele­ment is some­thing the com­put­er can’t imi­tate.

Capturing Thoughts and Feelings

The above sports sto­ry also includ­ed the quote from Paige: “I said jok­ing­ly to my team­mates that I was back.” An under­stand­ing of nat­ur­al lan­guage is and will con­tin­ue to be a very big chal­lenge for arti­fi­cial intel­li­gence.

We want to know peo­ple’s thought and feel­ings on the facts and stats. A robot can’t yet con­duct a nat­ur­al inter­view and fil­ter out the answers to be used as a key quote. It’s often in these quotes where the emo­tion of an arti­cle comes across.

Don’t Rely on Robot Writers Just Yet

Emo­tion, con­text, news sense and humor should all feed into a com­pelling con­tent cal­en­dar for brands too. You need to appre­ci­ate what your read­ers like, what mat­ters to them, how best to talk to them and how to enter­tain them.

There is a lot that hap­pens at Zaz­zle before we write any words on a page. This includes:

  • Dis­cov­er­ing the tar­get audi­ence of a brand and then cre­at­ing per­sonas
  • Key­word research to make sure we are tar­get­ing the rel­e­vant terms
  • Ensur­ing con­tent we plan to cre­ate is rel­e­vant and there is a vari­ety of it.
  • Under­stand­ing (and some­times cre­at­ing) the tone of voice
  • Once all this is done we will start to write. But, unlike a robot, we will be able to keep all of the above in mind while mak­ing sure that we write con­tent that oth­er humans want to read. Writ­ing involves ana­lyz­ing and inter­pret­ing this infor­ma­tion to deliv­er a mes­sage effec­tive­ly.

As Mankoff said: “Machines in the end are idiots, or maybe idiot savants, that need humans to cre­ate con­tent that’s going to be inter­est­ing to human beings.”