ODISSEI Lunchlezing: The Robot or the Brain? Building a Classifier for Visual News Frames of Artificial Intelligence

Op dinsdag 6 december, tussen 12:00 en 13:00, geeft dr. Irina Lock (Universiteit van Amsterdam) een online Lunch Lecture over het framen van Artificial Intelligence (AI) in online media afbeeldingen. In nieuwsmedia en publiekelijke online gesprekken worden digitale technologieën zoals machine learning applicaties vaak afgebeeld als magische “Artificial Intelligence” (AI). Nieuwsartikelen framen de AI-discussie vaak als baten versus lasten. Er is echter geen systematische kennis van de manier waarop AI wordt geframed. Om te ontdekken hoe AI wordt geframed in online media afbeeldingen, ontwikkelt Lock, in samenwerking met datawetenschappers, een machine learning classifier die automatisch online afbeeldingen categoriseert in voorgedefinieerde frames van AI.

Christiaan Meijer, research software engineer van het Netherlands eScience Center, die de classifier heeft geprogrammeerd, zal Lock’s lezing aanvullen met zijn presentatie over de technische uitvoering van het project.

We nodigen je van harte uit om deze online lezing bij te wonen en meer te weten te komen over Artificial Intelligence, machine learning en het doen van onderzoek naar visuele frames. Na de lezing is er ruimte voor vragen en discussie. Meer informatie over de lezing hieronder (Engels).

About this Lunch Lecture

In news media and online public discourse, digital technologies such as machine learning applications are often portrayed as magical “artificial intelligence” (AI). When technologies are novel and evolve rapidly, people tend to respond to them emotionally.

News and social media, and online communication environments more generally put such technological advances in a specific light through framing. News articles often frame AI’s risks versus benefits, and recent analyses point to an increasing, slightly positive, coverage of AI. However, analyses have so far focused on a story’s text only, neglecting the accompanying images. Visual framing studies have largely been using qualitative approaches and manual quantitative content analyses. This is because most news databases aggregate news texts without the accompanying images. Studies have also predominantly analysed news media, yet, not looked at online images more generally, for instance, user-generated content on public image databases, stock images used on corporate websites, etc.

This is a critical gap in our understanding as, particularly in a cluttered online environment, images direct the reader to content and gather their attention. As opposed to text, images evoke vivid associations to the real world, are more saliently processed, and more profoundly imprint in recipients’ memory and emotions. Given resource and methodological limitations, though, images have largely been neglected in content analyses projects.

Thus, the way AI is visually framed online – beyond news also on owned and shared media – purports multiple sociotechnical imaginaries of how the reader is supposed to envision the future and may ultimately influence the acceptance of this technology in society. Anecdotal evidence shows visualizations of AI repeatedly use stylized humanoid robots or brains, speaking to technology’s anthropomorphization. However, we miss systematic knowledge about how AI is visually framed online. Therefore, this project explores

Research Question: How is artificial intelligence framed in online media images?

To do so, it develops a machine learning classifier in collaboration with data scientists that automatically categorises online images in different pre-defined frames of AI. Thematically, this project is the first study to automatically analyse visual frames of this timely and relevant topic at a large scale. Importantly, this project is methodologically at the forefront in communication science and contributes to the growing community of computational communication scientists interested in analysing images. In doing so, it helps bring communication research forward not only by illustrating how computational methods can be used to detect and classify theoretically relevant concepts (i.e., frames) in images, but also by providing clear and tangible guidelines for how other researchers can apply these methods in their own work.

About Irina Lock

Irina Lock is Professor of Strategic Communication at the Institute of Communication Science, Friedrich-Schiller-University of Jena, Germany, where she researches the digital strategic communication of organisations and how organizations communicate around public issues such as sustainability. Irina is a former fellow of the University of Amsterdam’s Institute of Advanced Study (IAS) has been working at the Amsterdam School of communication Research (University of Amsterdam) as assistant professor. She has published on automated visual content analysis, digital issue debates, and the credibility of corporate social responsibility communication.

About Christiaan Meijer

Christiaan Meijer is an eScience Center’s Research Software Engineer with a background in Psychology, Physics and Artificial Intelligence. His specialties include machine learning, pattern recognition, computer vision and reinforcement learning. He was actively involved in programming of the classifier for framing of Artificial Intelligence (AI) in online media images, a project that was granted in the 2021 ODISSEI-eScience Call for proposals.


The ODISSEI Lunch Lectures highlight methodological issues and innovations in Social Science.

Photo by DeepMind on Unsplash