How AI in newsrooms supports good journalism in the Global South

Election debates can be rife with claims and counter claims – often presented as facts. Particularly in the speed of live debates, it is a challenge for newsrooms to determine what is correct, whether information has been taken out of context and the precise origins of the story data. Verification and factchecking are foundational building blocks of good journalism.

News outlet La Silla Vacia, an independent Colombian news outlet that covers politics and power in the country, has experimented with how to make factchecking more efficient. They have taken their first steps in bringing artificial intelligence into factchecking efforts thanks to a small Artificial Intelligence (AI) Impact Fund grant by IMS.

“These tools have already helped us to streamline a capital part of the factchecking process: we identify much more quickly the statements from political leaders that can be reviewed and eventually factchecked,” says Santiago Chavarriaga from La Silla Vacia.

The grant allowed La Silla Vacia to access and incorporate the bespoke solution Chequeabot, an AI bot developed by Chequeado, an Argentinean factchecking outlet. They have been using Chequeabot since last December in order to be more efficient and accurate in their daily production for the “Detector de Mentiras” (Factchecking area).

Chequeabot has three uses. First, it helps us to detect false claims in politicians’ speeches or tweets. It also records the speeches in order to facilitate the detection of false claims. Lastly, it detects false claims in pre-selected Twitter accounts.

The bot allows La Silla Vacia to have a daily customised queue of fact-checkable tweets by Colombian politicians. It also includes a tool that automatically selects fact-checkable sentences from a long text or a video (including transcribing).

The bot is manually linked to a database of previously factchecked material, allowing the factcheckers to quickly know if a claim has already been checked and reviewed. They want to progress with a tool that can automatically transcribe debates and select the claims that can be fact checked, and to have the AI review the database and produce a factcheck, when applicable, within seconds.

“We are using the Chequeabot on a daily basis and are able to spot three to five statements per week that end up being factchecked in a more timely fashion, and therefore our work can have much more impact on the day-to-day political debate,” says Santiago Chavarriaga from La Silla Vacia.

“We see the access to the bot as only the first step on the path to be able to factcheck in real time, or almost, an ability that can make the factchecking work much more powerful, both by providing information that can counter mis- and disinformation at the beginning of their lives and by being able to strike the iron when it is hot, to give the public a fact check when a debate is roaring.”

Faster factchecking during elections

Despite using it on a day-to-day basis, La Silla Vacia found the AI technology especially useful when reporting on the recent presidential election in Colombia in May 2022, when numerous claims were being uttered by several candidates in a short amount of time.

“Among other things, we used the new tools to factcheck the claims of a dozen presidential candidates in three separate TV debates, being able to publish all the factchecking within four hours of the end of each debate.”

To mention a few examples, La Silla Vacia was able to verify a claim that a new opinion poll had shown that a presidential candidate had taken the lead in Colombia’s Caribbean coast, a statement that was being hotly contested by followers and several candidates. The media factchecked the claim in under four hours from the moment the claim was originally stated. The AI technology also helped La Silla Vacia verify a claim of a presidential candidate that had stated that income from mining royalties was not being put to any use. Although verifying this required more work, they were able to factcheck the claim in only one day.

The Chequeabot saves La Silla Vacia a lot of time on the work they put into factchecking.

“We can now focus our efforts on reviewing and editing this automatic factchecking and researching the claims not previously factchecked,” says Santiago Chavarriaga.

He sees the implementation of the Chequeabot as a sign that AI can help promote good journalism if used in the right way.

“Our experience proves that technology can be an important ally of quality journalism even in a middle-income country and for a small, independent media outlet like ours,” he says.

The transformational potential of AI

The cooperation between IMS and La Silla Vacia is only one small part of IMS’ work with supporting media in the Global South in implementing AI solutions to promote good journalism and media business viability. IMS acknowledges that the transformational potential of AI integration in public interest media is immense. Still, it is evident that there is very unequal access to proper AI technology on a global level.  

The opportunities of AI technology in newsrooms exist at multiple levels of sophistication: from the use of single-application AI tools to automate or augment familiar tasks within existing journalistic workflows, producing familiar content items such as news articles or audio segments more efficiently to the use of entirely new AI-enabled workflows that have no manual parallel in newsrooms but which are still used to produce familiar content.

AI technology can also be used to effective implementation of paywalls, data mining, strengthening and scaling of investigative journalistic projects as well as automatisation.

AI mapping work by BBC News Labs suggests AI-augmented news production might become more stable, relevant and sustainable than existing forms of digital news, particularly where the focus is on synthetic media production. This includes natural language generation, text-to-image, text-to-video models, synthetic voices, synthetic avatars, speech-to-text models, automated translation, automated video editing and many others.

These AI and synthetic media technologies are augmenting human cognitive and communication abilities – and are not always used for good. They are now being used both individually and together to produce content experiences that rival or even exceed those produced by humans.

The possibilities of these new technologies are clear. However, the disruptive nature of AI and the lack of research and experience creates a learn-as-you-go approach and media outlets need to be aware not only of the possibilities of incorporating AI and algorithmic technologies but also of the dangers and the pitfalls. The digital technological “revolution”, now nearly two decades old, has necessitated a shift in normative media and organisational practices.

Bridging the global technology gap

In the Global South, especially in politically pressured and conflict contexts, there has been a tendency for a one-dimensional, almost instrumentalist view of new technologies and their impact in such spaces. Little attention is given to the contexts of their use and to what determines their adoption, appropriation, rejection, success or failure.

Broadly, discussions focus on generic issues such as information and communication technologies (ICT) access, digital divides, internet freedoms and internet connectivity, but they are reductive of a much more complex and textured reality.

“There are less visible yet fundamental factors that point to the need for a more intersectional approach to this subject. Such an approach will appreciate the plethora of interrelated factors that create the modes of simultaneous empowerment and disempowerment enabled by technology,” says Dr George Ogola, a leading researcher on the use of AI in Africa and part of the research team behind IMS’ forthcoming report on the use of AI in African newsrooms.

There are increasing global gaps between the Global North and the Global South when it comes to access and implementation of AI technology. Furthermore, a lot of the technology is developed in the Global North does not necessarily take into account its use in different journalistic and societal contexts.

“Gaps exist in the operationalisation of automated tools in many of the contexts where IMS works due to lack of data, skills and necessary management capabilities. Language barriers in the programming are a particular issue, and there are important considerations also in the development of AI to be reflective of racial and gender biases built into them. There is an urgent need to speed up our sectoral knowledge, so we can support media partners in this work – and make sure that they are not missing out on the potential of AI in promoting and sustaining good journalism in places where media and journalists are already under enormous pressure,” says IMS’ Head of Journalism and Media Viability, Clare Cook.

“The distinctiveness of the AI methodology at IMS turns the lens to intersectionality to better understand the requirements that determine the use and deployment of AI in newsrooms where our partners work and to the barriers hindering the adoption and use of AI at incremental and more extensive levels.”

How AI can be used to support public interest journalism

IMS will continue to work with partners on their efforts to use technology to strengthen their organisations and deliver public interest journalism, as well as advising and creating awareness of the risks and issues that come along with this.

IMS’ work with AI is a three-phase approach: mapping of AI use in different regions and contexts, training through workshops and supporting AI experiments in newsrooms.

“The first part of our work is to provide a detailed mapping in contexts where we work on the use of AI, and the knowledge around automation in newsrooms. Such a scoping study allows us to map the current state of AI in public interest media where we work, with context-specific understandings,” says Clare Cook.

As Dr George Ogola points out, this allows for more nuanced conceptualisation in the notion of public interest media and to operationalise artificial intelligence grounded in its situational use.

IMS examines how news organisations and journalists understand the notion of AI. Based on preliminary research, there is still great confusion over what AI actually means. Data and algorithmic literacy where we work remain relatively low and therefore likely to determine particular understandings and therefore use.

“We examine what and how AI is used in newsrooms: How are they changing or reconfiguring everyday newsroom practices across the continent? What are the professional and ethical issues around AI-driven automation processes in the context where they are used? What are the prevailing implications for AI-powered automation in a region so unequal in terms of resources, access, digital literacy and other divides?” says George Ogola.

Attention to context

As AI is primarily data-driven, IMS also looks at what data or algorithmic accountability challenges these technologies bring and how news organisations and journalists are responding. Methodologically, the project adopts a survey instrument, in-depth qualitative interviews of key stakeholders or informants and ethnographic newsroom observation.

“We pay particular attention to the broader contextual forces or factors that determine the deployment of and engagement with AI. Due to low data and algorithmic literacy, there is also little public participation, or ambivalence towards debates on critical issues such as data legislation and regulation,” says Clare Cook.

Generally, the conversations around data policy, legislation and regulation have been mainly initiated and sustained by civil society groups working with emerging young tech community. As such, data accountability mechanisms (legislation, policy and governance structures) are also only just emerging. Furthermore, many nations in the Global South have largely co-opted data and AI regulation, policy and strategy from North America and Europe. IMS explores the implications of this “borrowing” for the media sector to inform how it shapes newsroom practices.

Due to IMS’ large network of media partners around the globe, the organisation is in a unique position to make practical use of the research and help media experiment with AI solutions. This begins with regional workshops: 

“With local partners, we host a regional in-country AI challenge to include a training workshop and collaborative exploration of possibilities. The workshop is shaped by the findings of the first phase to ensure that it addresses the contextual needs of the region. The workshop seeks to address the knowledge gap on AI in newsrooms where we work and further explore ways of designing a contextually specific AI tool for the region’s newsrooms in which local editorial, user, technologists and policy input informs design. In the workshop, we also investigate ways in which such AI tools can be made both adaptable and scalable across the region,” says Clare Cook.

Finally, the third phase cements innovation in action. IMS’ runs an AI Impact Fund call for short-term AI experimentation for newsrooms. The transformational potential of AI integration in public interest media is immense. However, for such integration to be successful and to “smooth the transition” as public interest media in the continent embrace such disruptive new technologies as AI, research and experiments of this kind must be undertaken. The focus on the contextual specificities and therefore a holistic understanding of the needs of the local newsrooms will help avoid exacerbating current inequalities by simply transposing technology. The La Silla Vacia trial was one such experiment.

“Agility is critical in the sustainability of independent news. IMS’ approach builds on a fundamental belief that outcomes and key results in terms of what and how newsrooms learn is as important as concrete metrics and key performance indicators. The media development space has an important role to play in facilitating experimentation, even allowing newsrooms to fail forward,” says Clare Cook.

AI in emerging markets

IMS is committed to pushing back against two speed technological development in the Global North and South, with unequal access and knowledge to the ways AI technologies are put to use around the world. To gain a better understanding of how AI technologies are used in emerging markets and politically pressured contexts, IMS is leading on providing context-sensitive evaluations of the different possibilities and challenges when implementing AI in different countries and regions.

A first report, AI in Emerging Markets, qualified the conversation around AI in emerging markets of Latin America and Central Eastern Europe.

“We set out to shine a spotlight on the early AI adopters in the news media, not only newsrooms in the US and Europe, but also in Latin America and Central-Eastern Europe, where more and more publishers are facing up to ad revenue fears and experimenting with AI and new technologies as a digital catalyst to reinvent journalism. We wanted to create a more global, open and connected conversation,” says Robert Shaw, editor-in-chief of the report.

One of the findings in the report is that there are vast disparages to be found when comparing how AI technologies are used globally.

“The report shows worrying, but not surprising, tendencies,” says Jakub Parusinski, founder of The Fix, an organisation devoted to cracking the media management puzzle in times of massive transformations.

“New technologies, especially artificial intelligence and machine learning, are disrupting the world of media. But, as it often is, the future is not evenly distributed. There is a growing gulf between those able to make the digital leap and those stuck behind – a gulf with access and ability to use digital talent at its heart.”

But this does not mean that there is not plenty of interesting experimentation with AI technologies happening around the world.

“At the same time, we see that innovation is not limited to the so-called ‘Western world’. On the contrary, there is a lot of creativity and innovation happening in emerging markets. Making sure it is not happening in isolation is both a great challenge and great opportunity for the media development community,” Jakub Parusinski says.

AI in African newsrooms

A second study focuses on how, in a rapidly changing journalism ecosystem, news media organisations in Africa are increasingly adopting new digital technologies and business strategies to remain relevant, competitive and sustainable. The publication is expected to be published in December.

Across the African continent, many news media organisations are embracing digital first strategies, several positioning their digital platforms as their primary brands. Even where such digital shifts remain tentative, the changes are broadly acknowledged as critical to the continent’s media futures. 

This project seeks to map and understand the current state of AI use in African newsrooms and to explore its potential in strengthening public interest media in the region. The project will focus on country case studies across Eastern, Southern and West Africa. In Eastern Africa, focus will be on Tanzania, Somalia (especially Ergo), Ethiopia and Kenya. In Southern Africa the study will look at Zimbabwe, South Africa and Zambia, while in West Africa, Niger, Nigeria and Burkina Faso will be studied.

Adoption of digital technologies in African newsrooms varies greatly, with uptake much more visible in South Africa, Kenya and Nigeria, the largest media markets in the region, and less so elsewhere. In the three countries where there are also relatively well-resourced media sectors and actors, news organisations have adopted subscription models, introduced paywalls on their websites and begun distributing their content via platform media.

Some are now actively collecting personal data from subscribers and third-party vendors to understand their audiences better and to optimise the monetisation of their content as they deploy various technologies in their newsrooms. These processes, whether actively pursued, or simply because their enabling platforms employ some form of automation mean that artificial intelligence (AI) technologies are gradually becoming central to the (re)making of the African newsroom.

“As these AI tools and practices become commonplace in newsrooms, they have or will have significant implications for African journalism. They raise questions about their impact on everyday newsroom routines, journalism ethics, reliance on data largely controlled by an opaque infrastructure owned by Western big-tech companies and the appropriateness of the legal and policy infrastructure within which the news organisations operate,” says Dr George Ogola.

“How can they, for example, not merely end up reproducing the various inequalities that characterise many African societies? There is very little research on these emerging issues and how AI is being deployed in African newsrooms, much less their impact. This project will be among the first seminal attempts to critically engage with the subject.”