Technological development is responsible for enormous advances in interpersonal communication and access to information. Amid these positive points, an unexpected epidemic is emerging: infobesity, further promoting disinformation. Such a state of affairs is due to the overabundance of which the veracity is often doubtful, which makes decision-making in business complex.
Knowing that the detection of misleading elements by humans is often faulty, in this context there is an urgent need to develop a decision-making support tool. From this perspective, Andrés Romero’s dissertation “Artificial intelligence as a facilitator of decision-making” asks what IT solution can be implemented to facilitate decision-making in an environment of information manipulation? Could artificial intelligence be a solution?
The evolution of information
Having several meanings depending on professional circles, etymologically information comes from the Latin ‘informare’ which means “to give shape to”, “to form an idea of”. This assumes that information requires cognitive work to be named as such, since data is the raw element. More clearly, it constitutes “data to which a meaning (or an interpretation) has been added”.
Access to information as we know it today with hyper-sophisticated machines was not always done this way. From the earliest age of humanity, gestures and cave drawings were used as means of communication. Later, with the appearance of the printing machine in the 15th century, and other information media such as cinematography and the phonograph in the 19th century, its distribution became massive. This character will be amplified with the advent of the computer in the 20th century. Initially it will only be communication between humans and subsequently, between machines. This development has led to an explosion of information driven by the appearance of social networks, giving rise to the profession of big data, and raising the problem of its reliability.
The allusion to this characteristic brings to the fore the real manipulation of information through the use of marketing techniques in the advertising field, and through omnipresent digital surveillance, influencing users in their decision-making.
Under the influence of information manipulation
Decision making is driven by parameters of all kinds including emotions, habits, instinct, personality, information and many others. All these elements make human cognition limited, preventing it from making a completely rational decision.
Partial rationality during the decision-making process is blatant on social networks, serving personal interests. In May 2017, for example, an Australian newspaper demonstrated that Facebook took advantage of users’ vulnerability in order to influence product purchasing decisions by adolescents.
In view of this manipulation, measures are taken to detect it. This is the case of decoders from various reputable media who carry out “the recovery, compilation, explanation and classification of rumors. By correcting the facts, they explain to the public whether a rumor is true or not. “.
From the exit of the manipulation
In order to identify deception, artificial intelligence presents itself as a solution. Natural language processing (NLP) is one tool. It corresponds to all the language methods allowing a text to be transformed into a formal representation, in order to carry out calculations to find similar texts, extract precise information or identify information manipulation patterns. This is made possible thanks to the components of the language: syntax, morphology, lexicology, etc.
These components give rise to information analysis techniques: morphosyntactic labeling and lemmatization (attributing a neutral and generic form to a word). Although these techniques seem promising, many NLP applications are considered difficult to use and of little interest from a business point of view due to the ambiguity of human language.
Similar to NLP, web mining allows you to automatically discover and extract information from documents and services on the web. It aims to identify useful patterns in a set of text corpora. Which patterns will influence the decision-making of humans who must submit to the filtering process of web mining, in order to discern real information from false ones. Several linguistic cues make this complex operation possible:
- the relationship between words,
- the syllables,
- the meaning of the words.
We can also mention the grammar of fake news resulting in the use of grammatically correct sentences, and the simplicity of short tweets and Facebook posts. All this enhanced by arguments of authority like the institutions recognized for giving legitimacy to information. By computer, the index of deception is based on an absence of the contextual factor.
Experimentation and research results
The experimentation of the study reported in Andrés Romero’s dissertation is carried out in a startup Proleads. Romero’s main objective is to improve the quality of the company’s search engine results. As a first approach, it uses lexical analysis to identify the equivalence of lexical forms and in the second, web mining and the NLP technique knowing that a corpus of 3 million texts includes a high rate of presence of impertinent pages. From there it emerges:
- An increase in error page classification with accuracy up to 99.7%, or 15% of error pages identified compared to the previous approach.
- The contact page classification model has an accuracy of 98% with an increase of 10% in error pages identified compared to the previous approach.
Outside of the company, Romero explores web mining in situations of information manipulation. His choice concerns Covid-19 which has sparked debates both in society and in business. The idea here is to see to what extent the NLP technique accepting the addition of machine learning algorithms can simplify decision-making in this context.
Here the corpus studied is the decodex, the decoders of the World. After identifying and analyzing web resources talking about covid, and preprocessing the databases to recover the URLs, the researcher concludes that:
- NLP and machine learning facilitates the decision-making process by diagnosing information manipulation.
Reference
Andres Romero. Artificial intelligence as a facilitator of decision-making. Management and management. 2020.
Online https://dumas.ccsd.cnrs.fr/dumas-03000582v1
See more articles by this author
Related News :