During the first edition in Spanish of the Journalism AI Academy for Small Newsrooms, more than 20 journalists from small newsrooms from 15 Latin American countries will seek to learn how artificial intelligence can help their media outlets optimize processes, reduce the workload of their teams, better engage with audiences and improve sustainability.
For the first time, JournalismAI – the London School of Economics initiative for the responsible use of AI in news organizations – organized a Spanish version of its Academy for Small Newsrooms, a program that it has been carrying out since 2021 and that seeks to offer a deep dive into the potential that AI has for journalists and media professionals in newsrooms of less than 100 people.
Although JournalismAI has had initiatives for specific regions of the world, the Spanish Small Newsroom Academy is aimed only at participants from Latin America. For this first edition, participants from Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Peru, Uruguay and Venezuela were chosen.
¡Aplica antes del domingo a la Academia para Pequeñas Redacciones de #JournalismAI!
Esta vez, en español y enfocada en redacciones de #LATAM
— Polis@LSE (@PolisLSE) November 17, 2023
“This has always been part of the goal of the initiative of JournalismAI, to sort of bridge the knowledge around tech and AI in different areas of the world that are not just the Western, more developed countries. So this felt like a natural next step for us in developing this program,” Sabrina Argoub, JournalismAI Program Manager, told LatAm Journalism Review (LJR). “The programs and the cohorts that we have run, in order to cover the world, we just divided it, mostly for time zones reasons. And this year we want to experiment and try and go to a smaller region, and more focused on similar language and similar culture.”
The program, which began on Jan. 18, will last eight weeks and will feature renowned journalists from the region as instructors.
Some of the journalists who will participate in the Academy said they will seek to take advantage of the program to promote specific AI projects in their newsrooms, while others said they want to obtain skills to apply this technology in operational aspects of their media outlets, such as reporting processes, the dissemination of content and the strengthening of sustainability strategies.
Gianfranco Huamán, data journalist at the digital investigative journalism outlet Ojo Público, of Peru, said that among the ideas he has in mind is the application of generative AI to help his audience team generate content for social networks.
“Although it seems simple to generate tweets or copy for Facebook or Instagram, [the audience team] have told us that it takes them a long time to do that, due to the amount of news that is coming out, so what we have been thinking about is precisely to accelerate that process with artificial intelligence,” Huamán told LJR. “That is one of the goals that we want to achieve this year and that the Academy is obviously going to give us guidance on how to do it, if there are other colleagues who have already been doing it.”
Ojo Público has extensive experience in using innovative techniques to improve its editorial processes and investigations. In 2019, the media outlet developed Funes, a document analysis algorithm designed to detect risk indicators and possible traces of corruption in public contracts. Likewise, in 2023 the team presented Quispe Chequea, a tool that uses AI resources to produce verification content in different formats and in three Indigenous languages.
In addition, Ojo Público was one of the 40 media outlets that participated in the cross-border collaborative megaproject NarcoFiles, led by the OCCRP (Organized Crime and Corruption Reporting Project) and CLIP (Latin American Center for Investigative Journalism). For its part in the investigation – a report on the Venezuelan criminal gang Tren de Aragua– the Peruvian media outlet mapped the members of said criminal organization who had been captured in Peru in the last three years. They used AI to do this.
They used the API (application programming interface) of OpenAI, the organization that created ChatGPT, which gave them access to the engine of the popular application. With this and a Python code, they were able to analyze hundreds of journalistic articles and extract the data they needed more quickly.
“We had ChatGPT read this news and structure the data. The result was that he gave us a table that, although it was not 100 percent perfect – it had to be cleaned – in general terms, it generated very good results,” Huamán said. “It was a combination of traditional scrapping with ChatGPT language engine analysis, which I don't think would have been possible before, or would normally take a long time.”
The journalist hopes to continue using ChatGPT technology to optimize content creation for social networks. However, for the media outlet it is important that its texts on Facebook, Instagram and X are not only factually and grammatically correct, but that they follow Ojo Público’s style. And that is not achieved with the open version of ChatGPT.
“What, at least at the time of prototyping, we have been thinking about is doing a little fine-tuning, which is a technique in which you continue training the model. It's like a small adjustment, you feed it your own data so that the result of what you are looking for is better,” Huamán explained. “That is one of the methods we plan to use, but obviously we want to see if [at the Academy] there are other types of approaches or similar projects in which they have tried to address this problem.”
Another media outlet that has had significant experience in the use of AI for journalistic purposes is Cuestión Pública, from Colombia. It will be represented at the Academy for Small Newsrooms by Andrea Rincón, editor and investigative journalist.
Rincón said that the media outlet began experimenting with AI in tasks such as transcription and image generation for its gamified products, until last year they launched The Odin Project as part of their participation in the Artificial Intelligence Journalism Challenge (AIJC) of the Open Society Foundation. It is an AI tool that allows you to automate information from the databases that Cuestión Pública has created in previous investigations to produce new content in a faster and more efficient way.
"We have investigated, for example, more than half of the Senate, we have investigated the political houses of Colombia and we have a lot of information, so every time something happens with a character that we have already investigated, what Odin does is give us the information that we have about this character to come out in a timely manner,” Rincón told LJR.
The Odin Project (which takes its name from the acronym for Optimized Data Integration Network) won a special mention at the Splice Beta Journalism Festival in Chiang Mai, Thailand, in November 2023. Although it is currently in the testing stage in the Cuestion Pública newsroom, the media outlet seeks for the tool to help them make the most of their databases.
For Rincón, the outlet’s participation in the Academy for Small Newsrooms will be focused on acquiring the knowledge and tools to continue creating applications that help them optimize their processes and train their team of 20 people, most of them young, to confront the challenges faced by small, independent media with limited budgets.
But also, she said, the team will seek to acquire skills to improve the impact that the media outlet’s major investigations have on the audience.
“When you have data investigations, you have products that are so dense or so robust that segmenting them to audiences and making them reach an increasingly broader audience is an enormous challenge,” Rincón said. “I think there are many artificial intelligence tools that can facilitate that path to have more impact.”
Mijail Miranda, editorial director and product manager of the digital magazine Muy Waso, from Bolivia, is also among those selected for the Academy for Small Newsrooms in Spanish. He said he will also seek to learn how AI can help reduce the workload of his team, which consists of six people. However, his main objective is to acquire skills to integrate this technology into the workflows of the administrative area of the media outlet, especially in relation to sustainability.
“We had a very difficult year in terms of workload, so we want to reduce that load to avoid burnout and so that our team is calmer and has more time for more creative tasks, investigative tasks,” Miranda told LJR. “The other objective is that we believe that artificial intelligence can help us optimize some processes that have to do with our audiences, in terms of donations and communication services to third parties that we offer for our sustainability.”
According to Miranda, Muy Waso has had good results in its experimentation with generative AI in the automation of tasks such as transcription, writing copy for social networks and preparing non-journalistic articles. However, its attempt in 2023 to launch a chatbot to improve the user experience in its donation system was not entirely satisfactory.
“In the end we did not launch this chatbot because the different tests we did did not achieve the results we hoped for,” he said. “And it is from this frustration that we saw a good opportunity in this call, to be able to experiment, learn about new tools, see how far the capabilities of artificial intelligence can go in a newsroom and what benefits we could get from it.”
In 2023, Muy Waso began working on a training and education project for journalists and activists from various regions of the country on tools and methodologies against disinformation and hate speech, ahead of the next presidential elections in Bolivia, in 2025.
Miranda said that he will look to the JournalismAI program to learn strategies based on AI to optimize the training program, which they carry out in alliance with DW Akademie.
“The project is like an immersive and gamified experience. So I believe that artificial intelligence can help us a lot to manage the interaction with the people who participate in this training program,” the journalist said. “Not only chatbots, many artificial intelligence tools allow a degree of interaction that could help us strengthen this immersive and gamified look that we want to give to the project in general.”
Although this year is the first Academy for Small Newsrooms in Spanish, it is not the first time that journalists from Latin America have participated in JournalismAI training initiatives. In previous editions of the Academy and in other organization programs, such as the AI Fellowship Program, they have had a Latin American presence for the development of AI projects.
However, this year's participants agree that the fact that this program is offered entirely in Spanish and with instructors from the region is a good way to contribute to reducing the gap that exists between English-speaking countries and the rest of the world in relation to the development and use of AI.
“There are many tools that are in another language and many colleagues do not dare to use them simply because they do not understand or because all the tutorials are in English, and they simply give up. And that creates a big gap,” Miranda said. “Not all journalists in Bolivia or Latin America have an intermediate- or advanced-level of English. And it is also a reality that many institutions do not have the opportunities to launch programs in Spanish, but when they do, I celebrate it, because it is a principle of democratization and of bridging this gap a little."
JournalismAI recognizes that the application of AI has a cultural component, and therefore it is important to provide training programs that consider the language and culture of journalists from different regions of the world.
Argoub said that for the Spanish edition of the Academy for Small Newsrooms they focused on convening instructors from the region so that participants can have practical examples in Spanish and within the Latin American context.
"When we talk with our experts or participants in general throughout the programs, they often want to emphasize recognizing that AI is a cultural phenomenon as well, and it needs to be implemented within the culture," she said. “The way that this industry is different in Latin America compared to Europe or the U.S. it has an influence and how you choose and which kind of tools you use to implement it.”
The language barrier is not only evident when learning about AI tools, but also in their programming. Until the emergence of generative AI tools such as ChatGPT or Bard, which work in multiple languages, most applications were trained with information in English and with context from English-speaking countries.
“Language has been one of the great barriers to extending artificial intelligence tools in the region. And Latin America has very particular challenges that you cannot [address with AI] if you have tools in English, if you do not have the context,” Rincón said. “When you develop certain projects, certain products, certain languages that you have to train with your style, with your tone, with your editorial [line], well, it changes a lot from one language to another. I think that is also key to closing that bias, which is also so important in ethical terms.”
Este proyecto, en el que desarrollamos habilidades y adaptamos herramientas contra la desinformación y los discursos de odio en, tiene el apoyo y el acompañamiento de nuestrxs aliades de la @DW_Akademie_es.
La experiencia que ofrecemos es inmersiva/transmedia/gamificada. pic.twitter.com/Ve48hEtDtQ
— Muy Waso (@MuyWaso) January 12, 2024