“We must defend democratic algorithms and avoid succumbing to a data-centered approach, a ‘dataphilia'” — Interview with Professor Yves-François Le Coadic

(Yves-François Le Coadic) As Alexandra Vidal aptly pointed out during our conversation, artificial intelligence is radically transforming our professions and, in fact, all professions. The world of information is undergoing a radical shift. But our professions — librarians, documentalists, archivists, journalists, museum professionals, publishers — whose raw material is information, are the first to be impacted. It is not that doctors, lawyers, engineers, teachers, etc., are unaffected, but to a lesser degree. They certainly produce information, communicate it, and use it. But do they concern themselves with its validation, management, and preservation?

(Archivoz) You were one of the pioneers in defining Information Science as a social science. With the ubiquity of algorithms today, do you consider that we are still in an era of ‘social communication,’ or have we moved into an era of pure ‘data technology’?

(Yves-François Le Coadic) Right from your first question, the emphasis on algorithms and their ubiquity stands out. We ourselves have used them since the birth of information banks (which we, like computer scientists, called databases): Boolean algorithms, the famous formal rules “AND, OR”, and the corresponding software, DBMSs. Boolean logic has not disappeared in the era of AI; it has simply evolved, alongside other algorithms, into a more natural and flexible way of expressing the precision and purpose of research. While we were “flirting” with data, we always remained involved in the knowledge work process.

Unlike some artificial intelligence applications that use algorithms to calibrate knowledge in order to cultivate personalized beliefs, we must defend democratic algorithms. We must avoid succumbing to “dataphilia,” a data-centered approach. Data and data technologies have exerted a strong pull on the humanities and social sciences, even forging what I call a pseudo-discipline: digital humanities.

(Archivoz) In your work, the mediator’s character is central to satisfying the “information need”. In your opinion, can artificial intelligence truly play this role of mediation, or does it lack the human “intentionality” that you consider essential?

(Yves-François Le Coadic) In our professions, we are confronted today as we were yesterday — it is our credo — with the need to understand users’ information needs. When these needs are expressed correctly, there is no need to initiate an informational negotiation. But unfortunately, information needs are rarely expressed clearly; therefore, we should not expect the reader, visitor, etc., to know exactly what to ask.

It is up to the mediator to anticipate this uncertainty from the beginning of the interaction. As early as 1930, a manual of bibliographic services stated that the librarian had to be able to guess the unexpressed needs of people; they had to be able to “read minds”: “They had to have the art of discerning people’s unexpressed thoughts, which consisted in knowing how to give them what they do not know they desire”. In Plato’s Meno, in the episode of the slave, we find an illustration of the importance of this maieutic approach, exercised by the intermediary, which stimulates intellectual reflection.

So, can artificial intelligence discern people’s unexpressed thoughts? Can it respond automatically to natural human speech? My answer is no. An artificial intelligence can provide specific results and generic comments. But only humans can guide another person with nuance, understanding, and trust. Artificial intelligence can apply existing rules, but only humans can balance competing priorities when those rules conflict. We hope to establish a strong and effective informational interaction with artificial intelligence in a maieutic way. Is this the objective of the “educational” version of the AI “Claude,” with its new “Learning” mode, which helps students develop critical thinking skills through Socratic questioning? We also hope that it identifies our information needs and then provides us with the information we require.

(Archivoz) Information management has traditionally depended on the organization and preservation of meaning. In your opinion, does the massive integration of artificial intelligence into our management systems run the risk of transforming information into a mere statistical product, to the detriment of its contextual and historical value?

(Yves-François Le Coadic) With artificial intelligence (let us not forget it is a digital intelligence), we foresee the invasion of numbers and statistical and mathematical procedures. Here, once again, we must maintain lucidity. Since the 1920s, statistics and mathematics have been integrated into information science. And today, in the cultural, educational, and research sectors, the effective management of public services increasingly requires a wide range of tools adapted to cultural, educational, and scientific contexts, as well as to the size and nature of the service. These objectives are the same across the private, commercial, and industrial sectors.

These tools include those for analyzing the information needs of the community served, those for guidance and evaluation of operations, and those for measuring performance, providing institutions with a comprehensive set of performance indicators. They apply statistics and mathematics to books, articles, patents, libraries, documentation centers, archives, media, museums, and the Internet, based on new metric disciplines: bibliometrics, scientometrics, infometrics, mediametrics, museometrics, and webometrics. +1

Artificial intelligence also involves the use of mathematical procedures. It accumulates mathematical objects, such as algorithms that perform numerous logical, matrix, and probabilistic calculations. As the IEEE Computer Society recognizes, while artificial intelligence often outperforms humans in classification and pattern recognition tasks, it continues to face difficulties with complex tasks such as decision-making and resolving semantic ambiguities. These tasks, common in information management, require advanced forms of reasoning, such as common-sense and causal reasoning, which have not yet been fully replicated.

(Archivoz) For Archivoz readers, what would your message be to future information professionals? How can they preserve the human and social aspect of information flow in an increasingly automated world?

(Yves-François Le Coadic) Move from information to knowledge! Just as we witnessed the shift from the science of books, archives, objects, and media to information science, we must now consider shifting to knowledge science, which I do not confuse with cognitive science. Nor, especially, with communication science, whose excesses we all know too well. In France, we suffer from the coexistence of information science and communication science, resulting in a bi-science, “Information and Communication Sciences” (ICS), a regional science, an anomaly in the global scientific landscape.

Let us not forget that information is, in fact, knowledge recorded in written (printed or digital), oral, or audiovisual form on a spatiotemporal medium. We have focused much of our attention on technology and media, thus neglecting knowledge itself: a blatant example is information literacy in France, which resembles more of a “media literacy” approach, where the nature of the medium and the communication technique matter more than the meaning or content of the message. Certainly, we have focused on knowledge management and its application in decision-making processes. But, when using artificial intelligence techniques, what about the creation of knowledge, which raises particularly serious ethical questions?

Finally, let us not forget — as disinformation reminds us — that the knowledge we must transmit in our information systems must resist two realities. On one hand, common knowledge, opinions, and preconceived ideas. On the other hand, false scientific knowledge or pseudoscience. We saw this during the COVID-19 pandemic: a disastrous scientific publication policy led two major international medical journals to retract articles related to COVID-19; disastrous scientific communication campaigns originated within the scientific community itself; and, finally, a disastrous information epidemic, an infodemic. Mass media and a more or less specialized press gave the general public a wide audience for pseudoscience and disinformation.

But there is something even more worrying, and this must mobilize us: the desire to know, which is at the core of the need for information, is being challenged by populist movements that gain strength and encourage the rejection of knowledge. In a recent “Appeal for the Freedom of Knowledge,” it was reiterated that knowledge is a democratic power that some would prefer to see supplanted by the power of beliefs. The knowledge, the desire to know, is shared by many people; but the desire not to know, to turn away from knowledge, is also real. It manifests as rejecting rational arguments in favor of erroneous opinions and false knowledge. In the era of demagogues, the passion for ignorance triumphs (Mark Lilla).

Knowledge workers, such as information professionals, faced with these irrational frenzies, must be able to make a significant contribution, helping users take advantage of appropriate and timely knowledge resulting from their information searches. This also includes addressing ethical issues related to the management of private information.

We would like to close this dialogue by sincerely thanking Professor Yves-François Le Coadic. Having the opportunity to discuss the evolution of information under your attentive gaze is a privilege for Archivoz. We thank you for the time and dedication given to this interview, which will certainly serve as a reference for future discussions in our field!

Entrevistado