By Lucas Vieira de Araujo, professor of the Universidade Estadual de Londrina*
Mass communication was one of the areas most affected by the expansion of technology. Technological changes have also put the traditional media business model in check. In this context, technologies such as algorithms, artificial intelligence and Natural Language Generation (NLG) have emerged, which are increasingly dominant in media companies that use them for a variety of applications from news production to content distribution.
This growth, however, has had adverse effects. The search for gains at scale and agility in the processes generated criticism regarding ideological aspects, journalistic ethics and transparency. Studies have shown that computer systems used by many media organizations present aggressive prejudices and feelings as a result of standards-based operation
Despite the importance of the subject, the discussions in Brazil about the algorithms are still too early and isolated. This is one of the reasons why I tried to answer the following question: What do Brazil's largest national and regional communications companies think about algorithms, artificial intelligence and NLG, and what are the perspectives in the country on these technologies?
To answer this, I conducted a study with Brazil’s main communications groups at the national and regional levels: Grupo Globo; Grupo Folha; Grupo Estado; Grupo Abril; Sistema Brasileiro de Televisão (SBT); Grupo Record; Grupo RBS, Group RIC, and Emissoras Pioneiras.
Algorithms, Artificial Intelligence and NLG
Before dealing with the survey results, it is necessary to contextualize artificial intelligence, algorithms and the NLG language, since they are complex technologies. Algorithms are coded procedures for transforming input data into a desired output, based on specified calculations. The output can also operate as additional input for later algorithmic selection processes. This possibility is called the neural network, classified as an attempt to emulate the computational structure of neurons in the human brain.
In order to improve the degree of predictability, many neural networks and evolutionary algorithms are intertwined with artificial intelligence, a phenomenon classified as the ability of machines to develop intelligence similar to human intelligence. In addition to thinking and acting as humans, other requirements are needed from the machine endowed with artificial intelligence. One is the ability to use Natural Language Generation (NLG)to produce news stories. In media companies, this process occurs through algorithmic journalism, which employs algorithms, artificial intelligence, and NLG to generate news –from crime reports to earthquake alerts– with low human intervention, usually just a web address for the machine to create text.
Brazilian communications companies’ views on algorithms (Survey Question)
In general, managers of the Brazilian communications companies interviewed were skeptical of algorithms. They recognize that it is a promising technology with great potential for growth, but they do not see greater integration with the technology in the short term. For the CEO of Grupo Abril, Walter Longo:
"I think the use of algorithms is spectacular not to discover what people should consume but what the person wants to acquire. If I know right now, through the algorithms, what people would like to acquire, it's good for me to improve sales of the product or to attract a greater audience."
Respondents were concerned about the cost of technology versus the benefits it generates in terms of cost savings. Particularly in the broadcast TV sector, Brazilian communication companies see limited possibilities for adopting algorithms. For programming director of Grupo Record, Marcelo Caetano:
"Inevitably, whoever produces content will benefit from the algorithms. One example is the Netflix series House of Cards, which was created from decisions made by algorithms based on a database. However, we are 'groping' in this field in Brazil. We are not very clear how this interferes in the broadcast TV business."
In addition to the cost of the technology, respondents highlighted legal and regulatory issues as impediments. It should be stressed that the adoption of algorithms represents a change in journalists' attributions and in the very structure of newsroom positions.
A dominant factor that was little highlighted by the interviewees, but which international research has already proven, is cultural practice in the creation of news. The organizational culture of the main Brazilian media companies has not internalized mechanisms of integration between humans and machines for the production of news.
The executive editor of digital content at Grupo Estado, Luis Fernando Bovo, was the only one to affirm categorically that it would be positive for journalists to focus on less mechanical activities, such as fact finding, while the machine would collect data and produce less elaborate text: “I do not think it’s bad to have a machine producing news that is simpler.”
The fear of technologies such as algorithms, artificial intelligence and NLG may also involve other factors. Research has shown that it stems from human beings' fear of losing control of the situation, although they realize that technology often makes better decisions in certain circumstances. The manager’s hesitation is well founded. There is the possibility of errors, manipulation, commercial interests and politicians interfering in what the machine does. Studies have shown that human collaboration is essential to preventing failures, as well as to broaden the debate around mechanisms of algorithmic transparency.
One of the few unanimous opinions among companies was the perception that technologies such as algorithms, artificial intelligence and NLG harm society by producing and disseminating false news.
For the CEO of Grupo RBS, Eduardo Melzer, “algorithms are suitable for simple information and feedback. They are not appropriate for serious and professional journalistic production, which needs to have discernment, determination, plural vision and social responsibility.” The content, however, has not proved to be absolutely relevant. Research has shown that news organizations lack independent distribution networks. Dependent on search engines and digital social networks to generate audience and increase advertising revenue, media firms fail to show how necessary they are to society. Without that, they lose credibility, which completes the vicious circle of reduction of both audience and advertising revenue.
Algorithms, artificial intelligence and NLG have proven to be useful innovations for media companies primarily when they strengthen distribution networks, reduce costs, and improve engagement with readers/subscribers. These positive results, however, have not occurred in Brazil, but in companies that have been investing in innovations for years. My research has shown that this is due in part to media companies’ reduced interests in developing, testing, and applying innovations with algorithms, artificial intelligence, and NLG in the coming years.
What to expect for the future
Using algorithmic journalism is inevitable. The Brazilian managers all agree on this. The key is how to use this technology favorably. It may be developing systems of their own in which algorithms, artificial intelligence and NLG not only help to collect and select information, but also to produce texts. A change in mentality of Brazilian media company managers is important in this regard.
Another relevant change is to produce technology at low cost. To this end, it is imperative that media companies become more effectively involved in the development of innovation in the country. It is possible to create, develop and apply innovation without placing too much of a burden on the company since there is a greater exchange with the innovation ecosystem, especially universities, whose technical and applied knowledge can be useful in finding new paths.
 NLG is a subfield of Natural Language Processing (NLP) whose main characteristic is the ability to learn from itself, from iterative processes of trial and error. This skill, formerly restricted to humans because machines could not adapt to the vicissitudes of speech and text, is part of machines equipped with the capacity for learning with a large amount of data, as is the case of Machine Learning and NLP.
*The Knight Center occasionally publishes articles from researches about journalism in Latin America and the Caribbean. Lucas Vieira de Araujo is a Doctor in Communication from the Methodist University of São Paulo. He is professor at the State University of Londrina (UEL) and at the Faculdade Assis Gurgacz (FAG). E-mail: email@example.com. LinkedIn: Lucas Vieira de Araujo