“The computer will never make a story interesting”

• In "Fake and causes",

  20 Minutes

  sheds light on themes around conspiracy, fact-checking and issues for democracy.

• During a meeting on the link between technology and information during the event Médias en scène, we met Ioana Manolescu, information researcher at the National Institute for Research in Digital Science and Technology (Inria).

• “The world will never deprive itself of a journalist who is knowledgeable in a field, who has expertise.

  What I plead for is that from there we apply quick tools but that we must keep in the hand of someone who knows the field”.

Through its series of interviews "Fake and causes",

sheds light on the themes around conspiracy, fact-checking and issues for democracy.

gives the floor to researchers, associations, experts or other members of civil society to open the debate.

During the Media Day in the Seine, organized by

and France Info, we took advantage of a discussion on technology and information to meet Ioana Manolescu.

For nearly ten years, the computer science researcher, also a professor at the Ecole Polytechnique, has been trying to instil a healthier relationship between journalists and data processing.

You were one of the first researchers to take an interest in the ways in which technology and more particularly data processing could help verify facts.

Where did your observation come from?

I had made two observations simultaneously.

The first was that in France and more broadly in Europe, we have very high quality open databases built with taxpayers' money.

We have every right to access them and they contain very interesting information.

On the other hand, by reading the press daily, I realized that I was asking myself certain questions which in principle we could answer from available open data, but I did not see them addressed in the media.

Maybe you have an example?

During the 2008 crisis, I had heard the government promise that by investing a certain number of billions in the automobile industry, we were going to preserve jobs in the industry.

Maybe it was a really good idea, but I didn't have the answer.

A few years later, I said to myself that we had all the figures necessary to know if the investment had preserved jobs and how many.

However, in the press, no one questioned the evolution of employment in the automotive industry when we had all the data.

Is it a lack of awareness of the existence of data or an absence of free data available for use?

In the European democratic and technocratic sphere, open data is something very important.

In France - even if we are not the only country - we are lucky to have the interministerial department for digital technology which enormously encourages the creation of open data sets at the level of all the administrations of France, but also the development of small software bricks that allow specific analyzes to be carried out and which are made open source.

As a reminder, open source is free access programs.

It is often confused with open data, data that does not move.

Open source is more like code.

All these tools exist, the data too.

So why isn't anyone using them?

It was then that I learned that in many newsrooms,

You have since worked with

Our first approach is to ask journalists what they would like to use the data for.

We therefore worked on request by adding it to our existing model which is mainly used to exploit statistics.

Every IT project starts small.

Ours is getting relatively big and will still grow.

When we develop in a team of researchers, we look at how it works, we ask ourselves the question of what needs to be changed or if it continues like this.

It's a very iterative process.

IT and journalism do not necessarily use the same language.

How do you adapt your work to information professionals?

Usually, when I start working with journalists, I ask them what data they use.

There is something that a computer scientist does not understand right away, it is the notion of data quality.

Some of my scientific colleagues make verification systems that take “we ask Google” as a data source.

We look at the first 1,000 requests, take an average and select the most popular.

It's technically feasible, but it's worthless for journalists.

That's something we learned very quickly.

Journalists fear they will soon be replaced by artificial intelligence.

But is technology really so dangerous for the profession?

There is no perfect Artificial Intelligence (AI).

You should know that if you have a super-improved Artificial Intelligence, playing with it for two hours will inevitably make it do something stupid.

What we usually call "AI" are systems that have learned from a lot of examples.

Conversely, the examples that he has not seen, the AI ​​will not have learned them.

It is a disadvantage.

The second thing is that the AI ​​has no idea what it's talking about.

If we say “today the sky is…”, most AIs will complete with the word “blue”.

But they won't understand why if we don't explain it to them.

He doesn't understand the meaning of "sky", he just calculates probabilities.

Finally, the human will always need to be there to understand the need for information?

Of course, you always have to check by hand.

Each time we give a figure from the Insee database, we give the link to the page for double checking.

If we take the case of police officers who die doing their job in France, there are two different data: deaths on mission and deaths in service.

It's not the same statistic because “death in service” includes road accidents… which can really make a difference.

But to understand it, you need to have some expertise in the field.

The world will never be deprived of a journalist who is knowledgeable in a field, who has expertise.

The same human expertise will make it possible to identify the contacts and the right sources.

What I plead for,

it is that from there we apply quick tools but that we must keep in the hand of someone who knows the field.

The computer in general is not going to succeed in making a story interesting.

In recent years, we have noticed a growing number of

I believe the first step is to work with the right data sources.

In computing, we say “garbage in, garbage out”: if the data is wrong, nothing good will come out.

It is above all necessary to move towards the right sources of data and nothing else.

From there, you have to look at what questions you have to answer and what tool you put in place to check.

It is still up to the journalist to make the difference between what has been said and what is found in the sources.

By the Web

The fake news more relayed on Twitter after the purchase of the platform by Elon Musk

Company

"There is a very strong depoliticization of conspiracy" on the part of the government

• fake-off

• Media

• Computer science

• Interview

• Company

• Personal data

• Researchers