Fighting INFObesity
Scanlitt's triple therapy

A (relative) truth puzzle...

Information, beliefs, knowledge — the list of intellectual concepts is a long one. Add to it thought, language, fakes, bias, ultracrepidarianism¹, conspiracy thinking… Why this roll call worthy of Prévert? Because it perfectly captures what the COVID era so starkly revealed. What is our relationship with knowledge? When faced with the unknown, the instinctive questions are, in theory: What is this about? Who should we listen to? Whom — or what — can we trust in order to understand, to know, to judge? Yet this new informational universe seems to have made that exercise less a matter of difficulty than one of danger.

“During the COVID crisis, science was confronted with media and political distortions,” Isabelle recalls in her introduction. The sequence highlighted this old challenge summarised in the question asked by Pilate, the procurator of Judea: “What is truth?” Like any of us faced with a decision, he had partial, undoubtedly biased, information and little time. Not being certain that he had proven knowledge, he chose according to the belief that seemed most plausible, least implausible, at a given moment, in a given context.

Information from all directions: inspiration or suffocation?

The global internet and its sidekick, the ubiquitous smartphone, have brought about a Copernican revolution and opened the doors to the information society. Generative AI, in the same vein, has begun a quantum leap towards the knowledge society, a higher cognitive level, undoubtedly more noble. However, these oceans of information, which have also become bankable petabytes, are as much a source of hope as they are of venom. The deeper the abyss, the broader the field of intelligent inferences, but the more likely the intoxication of the depths, the loss of critical thinking, and the suffocation of neurons by information overload.

So what? Will the thickness of the layers of information and the associated processing technologies end up devouring their creators? Are humans, lost in a cognitive labyrinth that is constantly and uncontrollably expanding, heading “irreversibly but devilishly” towards a regression in access to accumulated scientific, philosophical, political and social wisdom, towards being swallowed up in the quicksand of information inflation?

The urgency of a crisis accelerates the response to the need for knowledge

Isabelle has been working to answer these questions for many years. From the beginning of her academic career, she focused on illiteracy; in the second half of her academic life, she focused on methods to aid knowledge creation.

« From the early days of the COVID crisis, doctors approached me. The public health emergency required them to quickly determine the truth about this virus... its origins, transmission, diagnosis, treatment... except that a tsunami of publications was flooding in, closely followed by a tsunami of patients... » says Isabelle, explaining the challenge faced by the medical community and, consequently, by the politicians at the helm. With lives at stake, how could they find the needle of truth in the enormous haystack, a tangle of deceptions, fantasies and seriousness?

She then created a registered charity. Her team set to work with two objectives. First, to complete the automation and development of the tool for searching, selecting and classifying the mass of scientifically validated publications relevant to a given research topic. Then, to train the right AI model, calibrated to synthesise in natural language the publications selected and classified in the previous step. The approach, while not simple to design, has the merit of being based on a foolproof logic. The quality of a library is measured by the size of its document collection, but even more so by its metadata classification system, which allows users to quickly access relevant works. Let's call it a “meta-library”. Scanlitt went on to draw inspiration from this.

Successful transformation from association to start-up

Isabelle and her colleagues initially sought to build on existing metadata bases, which proved to be of poor quality in terms of content and services. As no one can serve you better than yourself, this group set about creating its own bases by qualifying the data to avoid garbage in, garbage out. After the COVID sequence, the decision was made to transform the association into a start-up. Scanlitt was born. It received every possible blessing from Bpifrance (labelled deeptech, frenchtech émergence, France 2030, etc.) and Europe (WomenTech EU) ... Fairy godmothers waved their magic wands. It filed patents and adopted an eloquent slogan: “Explore scientific literature at the speed of light”... a beacon shining in the heart of the storm!

DATA4S, storing high-quality metadata

In 2025, our body of knowledge is a virtual, global but disparate library, consisting of hundreds of databases containing (almost) all publications² produced since the advent of writing. Scanlitt enthusiastically brought together part of this universe within a data warehouse. With the support of FNEGE³, the start-up began to set up its metadata database, now called DATA4S (DATA for SCIENCE), and continues to feed it with daily updates. A directory of more than 30 million publications that continues to grow, it is currently divided into three entities: DataSoc (social sciences and business), DataHealth (health sciences) and DataEngi (engineering sciences), which is currently being set up.

ARTIREV, mapping research, prioritising sources

Named ARTIREV, this first software tool designed by Scanlitt is “deterministic”, insists Isabelle. ARTIREV works on the basis of DATA4S metadata⁴. It selects and summarises scientific publications relevant to the subject expressed in the user's query. As a deterministic algorithm, ARTIREV produces, without any possibility of error and in a matter of seconds, an elegant and eloquent map of the field being explored, then prioritises the results to define the researcher's reading priorities. At this stage, the researcher knows exactly where to find the publications that they absolutely must explore.

SOCRATE: an agent for synthesis and interpretation

Scanlitt's second tool, called SOCRATE, is a probabilistic generative AI agent designed to help users effectively exploit the results provided by ARTIREV: structuring the field of research, generating summaries in natural language, and assisting with interpretation by identifying relationships between publications. ‘SOCRATE cannot hallucinate (in the AI sense), because it is powered by ARTIREV, a deterministic tool. Our tools improve researchers' efficiency by an average of 60%,’ Isabelle insists.

Tech for Humanity...

“The real challenge for Scanlitt lies in ensuring the quality of its metadata bases,” Isabelle summarises. Scanlitt naturally offers its services to its customers but also seeks to promote ‘open science’. DATA4S is therefore made available free of charge to users of the platform. « Faced with scientific information overload, we want to provide the right tools and best practices, all in a transparent manner. Science must help decision-makers and citizens, not the other way around, » she comments.

Scanlitt does not rule out expanding into other areas such as the legal system. “Scanlitt optimises our entire relationship with knowledge. Access to scientific discoveries is a right for everyone, and also a matter of mutual respect. I aspire to an intellectual democracy, where decision-makers and citizens are informed, rely on reliable knowledge and use their brains,” concludes Isabelle. She adds, mischievously, that some time ago, the WHO announced with a high degree of probability that a new pandemic caused by a mosquito was imminent. “Let's learn from the previous pandemic! I call on our governments to support us so that we have the right tools and data when the time comes. I also invite institutions such as INRIA⁵, INIST⁶ (CNRS) and the CNRS itself to join us in helping us continue to gather the necessary data," she says, as a call to action at the end of the interview.

At Scanlitt, knowledge is first and foremost a “common good”. Its promoters do good. Its polluters, monetisers and impostors, on the other hand, do a lot of harm. The excesses of the internet have been a concern for its detractors since its inception. Combined with those of AI, they can further disrupt ethics. Scanlitt demonstrates, on the contrary, how the power of these technologies, when properly implemented, can not only correct their missteps, but also enlighten and provide major benefits to their users.

In a recent publication, Étienne Klein quotes an anagram by Olivier Garcia that Scanlitt could adopt as its own: “la dévalorisation du réel” (the devaluation of reality) gives “trouvaille de la déraison” (the discovery of unreason). The words shed light on our anxieties with their coincidences. What lucidity! Let us add that Knowledge goes hand in hand with Confidence, and Scanlitt with Success.

1. This term became popular during the COVID crisis. It refers to the act of giving one's opinion on everything without having any proven knowledge or expertise on the subjects in question.

2. Periodicals are listed by their ISSN (International Standard Serial Number), publications by their DOI (Digital Object Identifier).

3. FNEGE: National Foundation for Business Management Education.

4. The title of the publication, of course, but also the author, journal, date, keywords, abstract, etc.

5. INRIA: National Institute for Research in Digital Science and Technology.

6. INIST: Organisation affiliated with the CNRS, Institute for Scientific and Technical Information (digital information and services for research).