AI In Journalism

Share on facebook
Share on twitter
Share on pinterest

AI in the newsroom

Each​ ​newsroom​ ​has​ ​a​ ​unique​ ​set​ ​of​ ​ways​ ​it​ ​uses​ ​AI.​ ​For​ ​the​ ​first ​ ​session,​  participants​ were​​ asked​ to​ ​reflect​ ​on​ ​the​ ​role ​ that​​ ​artificial ​ intelligence​ ​currently​ ​plays​ ​in​ ​their​ ​newsrooms,​ ​and​ ​the ​ issues​ they ​ are​ ​ confronting.​ Chase ​ Davis,​ editor​ of  interactive​ ​news​ ​​at The New York Times​,​  ​highlighted  the​ ​ongoing ​ promises around  the​ melding ​​of​ ​technology​ ​and journalism​ to ​ help​ ​ reporters find​ and ​ tell​ ​ stories​ ​ that​ were previously ​out​ ​of​ ​reach​ ​or  ​​impractical.  


Areas where AI is particularly helpful can be classified into three categories:

1.Finding​ ​needles​ ​in​ ​haystacks:

​ In​ ​ ​those​ ​outlying​ ​or​ special​​ cases​ ​ that​ ​ might​ ​ elude​ ​ human​ identification ​ because​ of​ the​ ​ scale​ or​​ ​complexity​ ​of​ ​the​ ​data, ​ ​AI​ can​ be​​ a​​ breakthrough​ tool. ​ ​This​ ​role​ ​fits​ ​neatly​ ​into​ ​standard​ newsroom​ processes, ​ because​ even​ ​if​ ​it​ ​discovers cases ​ the human​ eye​ could​ ​ ​not, the​​ ​findings ​can ​be ​fact-checked ​ via​ standard​​ ​human investigative ​ techniques. ​

2.Identifying​ ​trends​ (​ or​ ​departures​ ​from​ ​trends): ​

The​ ​massive​ ​computing ​ power​​ AI​ can help ​ provide​ ​ characterizations​ ​of​ ​aggregates​ ​of​ ​data, ​ ​perhaps​ grouped​​ ​in ​ time​ or​ by​ geography​ ​or​ ​demographics. ​ ​Alternatively, ​ ​it​ ​can​ ​quickly​ ​identify​ ​outlier​ ​data.

3.Examining​ ​an​ ​application​ ​of​ ​AI​ ​or​ ​computation​ as​ ​​the ​ ​subject​ ​of ​​the​ ​story​ ​itself:

Because ​ they​ are​ ​ built​ ​ by​​ humans, algorithms​ ​harbor​ ​human​ ​bias—and​ ​by​ ​examining​ them, ​ we​ ​ can​​ discover​ previously​​ unseen​​ bias. ​How​ ​ are​​ these​ ​complex​ ​​truths​ being​ found ​ through​ ​these​ ​tools? ​ ​What​ ​happens​ ​when​ these​​ tools​​ ​are​ ​applied​ ​to​ ​the ​​operation of​ ​our​ ​neighborhoods​ ​or​ ​cities​ ​or​ ​nation? 


As​ ​suggested​ ​by​ ​Mark​ ​Hansen, ​ ​we​ ​are​ ​also​ ​starting​ ​to​ ​see​ ​situations​ in​ ​ ​which more​ ​ advanced​ journalists​ ​are​ creating​ ​ analysis​​ methods that​ essentially​ ​cultivate computational​ sources​ ​ on​ a​ given​ ​topic. ​ ​In​ ​those​ ​cases, ​ ​we​ ​have​ new​​ hybrid​ ​​forms​ ​of​ ​investigation​ ​and​ ​writing​ where​​ ​the story​ ​and​ ​the​ ​mechanistic​ ​technique​ ​share​ ​the​ ​spotlight.

A​ ​Spectrum​ ​of​ ​Autonomy

The​ ​incorporation​ ​of​ ​AI​ ​into​ ​the​ ​newsroom​ ​has​ led​ to​ a​ ​ significant breakthrough​ in​ ​ ​the​ abilities​​ ​of reporters ​ to act​ as amateur​ ​data​ ​scientist. ​AI can augment ​​the ​human ​reporter ​in​ several​ ​ways: ​ 

i)helping ​ to classify and​​ categorize​ ​documents,  

ii)identify​ outliers ​​in​ ​data​ ​worthy​ ​of​ closer​ examination  

iii)find​​ needles​ in​ ​ the​​ haystacks​​ of​ ​​data. ​ 

iv)keeping​ ​an​ ​experienced ​​human in ​ ​the​ ​loop​ ​with​ ​real​ ​news ​judgment​ ​ was​ frequently​ ​  referenced​ during​  ​ our​ discussion​​ as​ an​ essential​ ​part​ ​of​ ​working​ ​with​ ​AI​ in​ ​ the​ newsroom.​ One​ panelist​ ​​pointed ​​out​ ​there​ ​is​ ​a​ ​“spectrum of ​ autonomy”​ with​ respect to​ AI:​ ​​at​ ​one​ ​end​ ​of​ ​the​ ​spectrum​ ​is​ ​full​ ​autonomy,​ ​where ​​no​ ​human​ is​ in ​ the​ loop​ at​ all,​ ​and​ at​ the​​ other​ ​end,​ AI can​ work​ ​alongside​ ​humans in​ ​a​​ ​much more​ ​limited​ way.

 AI​​ ​can help​ ​free writers from​ ​​having ​ ​to constantly ​re-write​ ​the same stories over​​ and​ over​ to​ ​​work​ ​on​ ​more​ ​original​ ​reporting, ​ ​as​ ​long​ ​as​ humans​​ ​are​ ​helming​ the​ operation ​ and​verifying outcomes. ​  

 Major examples include​ Los Angeles Times reporters ​ using ​ classifiers ​to detect​ instances​​ of​ the​ ​LAPD​ ​downgrading​ ​crime​ ​classifications;​ The Atlanta Journal-Constitution​​ investigation​ ​of sexual ​ abuse​ by​ doctors;​​ Reuters’​topic ​ ​modeling to​ find​ centers​  ​of ​power among​ petitioners​ of​ the Supreme​​ ​Court;​ ​ProPublica​ ​demonstrating​ how machines ​​learn​ to​​ be​ ​ racist​​; ​and​ The New York Times​​ ​interrogating​ campaign​ ​finance ​ ​data and​ using​ facial ​ recognition​ ​to determine​ ​who was​ ​sitting​ ​in​ ​the​ ​audience​ ​at​ ​President​ ​Trump’s​ ​inauguration.


How​ ​does​ ​technology​ ​fit​ ​in​ ​the​ ​news​ ​pipeline? ​ ​As​ ​mentioned​ ​earlier, ​ ​AI​ ​increasingly​ ​assists​ ​in reporting, content creation, distribution, ​​and​ ​audience​ ​interaction, to​ ​name​ ​a few​​ ​examples. Recently, ​ crowdsourcing, brainstorming, and​​ fact-checking​​ tools​​ ​are being developed ​to ​aid​ ​data information ​ gathering​ and, particularly, to​​ structure​​ ​relevant ​data. Among​​ ​contemporary newsrooms, ​ automation​​ is​ ​ a​ ​ key​tool​ ​ ​in ​ competing​ ​ not​​ just against​ ​ ​each​ ​other​ ​for​ ​customer attention, but ​also​ against​ large platforms​ such ​ as​ Netflix, Facebook, ​ and​ ​Amazon. ​ The number of articles written by robots is growing rapidly. Just input the facts and let the algorithm write the paper. Associated press creates more than 3,000 financial reports per quarter. Some estimate that 90 percent of news could be algorithmically generated by mid-2020s

The above chart describes how software and journalists are perceived. As you can see, software totally dominates over the journalists.

An example of the use of artificial intelligence in journalism

Used by the Washington post after it was acquired by Jeff Bezos. The editors there started using Heliograf.

For an area of interest:

i)Hook heliograf up to source of structured data eg election data from

ii)Create narrative templates for stories, including key phrases that account for different potential outcomes

iii)Heliograf identifies relevant data, matches it to corresponding phrases in template, and then publishes different versions across different platforms.


In November 2012,

 It took four employees 25 hours to compile and post just a fraction of election results manually

In November 2016,

Heliograf created more than 500 articles, with little human intervention, that had more than 500,000 clicks


Use Heliograf to keep data in stories up-to-date

E.g., someone shares a Tuesday story on Thursday, and the facts have changed, Heliograf automatically updates story with most recent facts