AI and the media: a (r)evolution in investigative reporting?

Arti­fi­cial Intel­li­gence (AI) is every­where, gradu­ally find­ing it’s way into the inform­a­tion sec­tor. Could AI be the next revolu­tion in the media? The answer is more complex…

For the past ten years or so, the online press has been test­ing gen­er­at­ive AI to auto­mat­ic­ally write art­icles by “suck­ing” inform­a­tion from reli­able data­bases. In March 2015, before Chat­G­PT, Le Monde news­pa­per had used AI developed by the com­pany Syl­labs to write 36,000 art­icles cov­er­ing the res­ults of the region­al elec­tions in each com­mune, thanks to data from the Min­istry of the Interi­or. A few months later, France Bleu fol­lowed suit, this time for the region­al elec­tions. Since 2021, L’Équipe has also been using AI to auto­mate the pro­duc­tion of con­tent list­ing upcom­ing matches, fol­lowed by the time and chan­nel of broad­cast. Are journ­al­ists being replaced by robots?

“Arti­fi­cial intel­li­gence is first and fore­most a tool” says Ioana Man­oles­cu, a com­puter sci­ence research­er at Inria spe­cial­ising in large-scale data pro­cessing and fact-check­ing. “In edit­or­i­al depart­ments, there is a mix­ture of mis­trust and fas­cin­a­tion” she con­tin­ues. “While text gen­er­a­tion works well with Amer­ic­an AI – which is more highly trained – French AI is still rather clumsy.” What’s more, the abil­ity to ana­lyse, inter­view, cross-check sources and carry out in-depth invest­ig­a­tions remain exclus­ively human skills for the time being. So, there is no risk of journ­al­ists being replaced by “robot edit­ors” for day-to-day work, as some people envis­aged a dec­ade ago.

“On the oth­er hand, AI is very good at retriev­ing, organ­ising or com­par­ing con­tent with an exist­ing data­base” explains Ioana Man­oles­cu. This is the ori­gin of the Statcheck tool, which the research­er and her team have been devel­op­ing in col­lab­or­a­tion with Radio France since 2021. This AI allows stat­ist­ic­al inform­a­tion found in an art­icle to be cross-ref­er­enced with the INSEE data­base and, more recently, the Euro­stat database.

“To do this, we had to train the AI on a lot of texts to enable it to learn, for example, that the con­cepts of “employ­ment” and “unem­ploy­ment” are related…” explains the sci­ent­ist. Today, StatCheck can recon­cile the way journ­al­ists write with the ter­min­o­logy used by stat­ist­i­cians. New func­tions con­tin­ue to be added by the Inria teams, such as open­ing the sys­tem to extern­al data­bases, then stand­ard­ising sources in a single format. But the aim is to go even further…

It all star­ted with an invest­ig­a­tion car­ried out by a journ­al­ist from Le Monde a few years ago, who combed through hun­dreds of doc­u­ments to identi­fy the links between sci­ent­ists and indus­tri­al lob­bies… all by hand. “It’s a colossal job!” com­ments Ioana Man­oles­cu. So, the idea was born to cre­ate a tool to auto­mate the work. Called Con­nec­tion­Lens, it is now cap­able of inter­con­nect­ing data taken from a highly het­ero­gen­eous cor­pus of doc­u­ments (PDF, Excel, URL, etc.). “The AI extracts the inform­a­tion con­tained in these doc­u­ments, such as names, organ­isa­tions, dates, e‑mails, etc., and links them togeth­er. The name of a com­pany in the acknow­ledge­ments of a thes­is, for example, will be linked to the name of the author.”

But the algorithm is nev­er the last link in the chain! For both StatCheck and Con­nec­tion­Lens, a journ­al­ist sys­tem­at­ic­ally re-checks the work and the sources. “In my opin­ion, the only thing you shouldn’t try to ask the AI to do is to think” con­cludes the researcher.

Le Monde’s Décodeurs teams have adop­ted this approach in their work. “AI is a very good way of sift­ing through inform­a­tion and sav­ing time, by going from 3,000 names, for example, to the 200 that interest us” explains Maxime Vaudano, who coordin­ates the Les Décodeurs invest­ig­at­ive unit. The journ­al­ist, who spe­cial­ises in open-source and col­lab­or­at­ive invest­ig­a­tions, admits that they are not “very organ­ised or very impress­ive at this stage” when it comes to edit­or­i­al work.

AI is a very good way of sift­ing through inform­a­tion and sav­ing time.

Indeed, although sev­er­al large-scale invest­ig­a­tions involving very large data­bases, such as the Panama Papers, have already been car­ried out in the past using more basic algorithms, this does not guar­an­tee the reli­ab­il­ity of AI today. “We keep in mind that it is an imper­fect tool! There are, of course, false pos­it­ives, which are pieces of inform­a­tion kept by the AI when they are not rel­ev­ant, but the biggest risk remains false neg­at­ives” explains Maxime Vaudano. “When a piece of inform­a­tion falls by the way­side even though it’s rel­ev­ant to our subject.”

So, the use of AI is far from sys­tem­at­ic, and sev­er­al “clas­sic” invest­ig­at­ive tech­niques are reg­u­larly used in par­al­lel. “We are still involved in very long-term research, with a huge num­ber of veri­fic­a­tion steps.” So, whatever the ini­tial meth­od used to obtain the inform­a­tion, it will be checked sev­er­al times… by humans.

Is this then a case of simple evol­u­tion rather than revolu­tion? That’s cer­tainly the view of Ioana Man­oles­cu, who remains very cau­tious about the real cap­ab­il­it­ies of AI, but also about those of the hypo­thet­ic­al AGI (Arti­fi­cial Gen­er­al Intel­li­gence) of the future. “Yes, com­puters can beat humans at chess, so in that respect it’s all over. But we’re talk­ing here about a very spe­cif­ic sys­tem! As for the rest, “intel­li­gent” robots don’t know that water is wet or that time only flows in one dir­ec­tion, where­as a baby does…”

As for AI as an invest­ig­at­ive tool, the research­er is more optim­ist­ic. “I tell myself that young journ­al­ists will be able to learn these tech­niques and bring them into their news­rooms.” Because the use of AI remains linked to a cer­tain tech­no­lo­gic­al cul­ture with­in the media, but also to the train­ing avail­able to journ­al­ists to get to grips with these systems.

“We don’t have enough time to put these tools in place, even if the tech­no­logy is already avail­able” adds Maxime Vaudano. But the situ­ation is chan­ging. Since 2023, Report­ers sans frontières has been devel­op­ing an AI pro­to­type for journ­al­ists, ded­ic­ated to envir­on­ment­al issues and news veri­fic­a­tion. It has been under­go­ing tri­als since April 2024 with 12 part­ner media out­lets. At the same time, in March 2024 the news­pa­per Le Monde entered into a fin­an­cial part­ner­ship with OpenAI, the par­ent com­pany of Chat­G­PT, to give their AI access to their archives. This alli­ance is being care­fully mon­itored by teams of journ­al­ists, but it her­alds an unpre­ced­en­ted part­ner­ship between the media and intel­li­gent technologies.

Sophie Podevin