Artificial intelligence in the information sciences: meeting points and limits
Not a single field of activity escapes the influence of artificial intelligence (AI), but distinguishing its contribution and its limits is sometimes a challenge. In the field of information science, AI touches every stage of the information cycle, with varying degrees of impact. Cogniges offers a brief overview of the issue. To do so, we define the concepts, discuss the promising functions and challenges of artificial intelligence at each stage, and conclude with the elements that experts believe will require the intervention of information specialists for a long time to come.
The concepts involved
Definitions already show the potential areas of convergence.
Information sciences "study the properties of information, the forces that govern its flow, and the means of managing it to optimize access and use. They are concerned with the creation, collection, analysis, organization, evaluation, dissemination, transformation and use of recorded information and knowledge in all its forms." [1]
Artificial intelligence refers to "any tool used by a machine to reproduce human-related behaviors, such as reasoning, planning and creativity. More specifically, the European Commission considers AI to include: 1) machine learning approaches; 2) logic and knowledge-based approaches; and 3) statistical approaches, Bayesian estimation, and search and optimization methods." [2]
Information gathering made easy
Considering that the volume of information is growing daily, that sources are diversifying and that the media are more heterogeneous than ever [3] :
- Artificial intelligence tools such as ChatGPT, WebChatGPT or Perplexity can help to identify the semantic field relating to a literature research or information monitoring project [4].
- In the same way, they can help identify the most frequently used keywords, as well as the names of the main players or sources to follow in a field.
Large-scale processing, analysis and organization
- One of the strengths of artificial intelligence lies in its ability to perform large-scale processing on corpora that would be difficult to master using a human approach alone [5]. For example, it can automatically classify sources according to their lexical field, their referencing and the recency of their content.
- In a context of information overload, artificial intelligence can be put to good use to extract named entities (people or organizations), concepts and themes, since everything that is written becomes exploitable in natural language, and therefore a potential object of query [6].
Two examples are reported in an issue of I2D: information, données & documents devoted to artificial intelligence [7, 8]:
- Every year, 1 million new articles are indexed in Medline, whose thesaurus includes 27,000 descriptors (MeSH). The introduction of the Medical Text Indexer, which incorporates deep learning approaches, makes it possible to meet the need for rapid access to information by performing automated indexing controlled by human curation [7].
- Estimating the value and impact of a scientific publication is a challenge. Researchers have developed the MyScienceWork application to analyze bibliographic citations in large corpora according to three models: 1) Does the citation confirm or invalidate the author's thesis? 2) Does the author express a positive, negative or neutral opinion of the citation? 3) Does the bibliographic reference appear in the status, methodology or results? [8]
More targeted information dissemination on the horizon
Artificial intelligence is still in its infancy in this field, but it looks promising [4] :
- We can anticipate the targeting of intelligence information according to the profile of the recipient.
- We can also foresee the possibility of structuring and presenting information according not only to the recipient's profile, but also his or her level of specialization, language register and reading habits.
Roles still devolved to information specialists
- A thorough understanding of the customer's context and needs remains the prerogative of well-trained human beings. At Cogniges, we see this every day.
- For the time being, only specialists in the relevant fields or in information are in a position to assess the quality of sources [6], carry out quality control on results and guarantee the appropriate use of artificial intelligence in line with ethical standards, laws and regulations [3].
- If artificial intelligence can take over repetitive tasks, the information specialist can concentrate his or her efforts on higher value-added tasks, such as handling difficult cases and making complex decisions.
- Writing a synthesis - not a summary - is a combination of a specialist's understanding not only of the content, but also of the specific needs of his or her customers [4, 6].
While the contributions of artificial intelligence have the potential to free up information specialists' time for project management and complex, value-added tasks, they nevertheless require the development of new skills and the mastery of new tools, such as those dedicated to exploiting metadata, or feeding and perfecting algorithms [3].
What's more, over and above the need for government authorities to regulate the use of artificial intelligence, information specialists will always be using their ability to adapt, their critical thinking, their emotional intelligence and their ethics in the service of their customers' information retrieval and monitoring activities: components that will not be integrated into an algorithm any time soon, no matter how sophisticated it may be.
References
[1] Université de Montréal. École de bibliothéconomie et des sciences de l’information. (s.d.). Les sciences de l’information.
[2] Commission nationale de l’informatique et des libertés (CNIL, France). Glossaire de l’intelligence artificielle.
[3] Jacob S., Souissi, S. et Martineau, C. (2022). Intelligence artificielle et transformation des métiers de la gestion documentaire. Québec : Université Laval, Chaire de recherche sur l’administration publique à l’ère numérique, 17 p.
[4] Deschamps, C. (2023, janvier). Ce que ChatGPT fait à la veille [4 billets]. Outils froids.
[5] Chartron, G. et Raulin, A. (2022). L’intelligence artificielle dans le secteur de l’information et de la documentation : défis, impacts et perspectives – Présentation du dossier. I2D : information, données & documents, 1(1), 8-12.