The browser you are using is not supported by this website. All versions of Internet Explorer are no longer supported, either by us or Microsoft (read more here: https://www.microsoft.com/en-us/microsoft-365/windows/end-of-ie-support).

Please use a modern browser to fully experience our website, such as the newest versions of Edge, Chrome, Firefox or Safari etc.

New doctoral thesis: How scientists teach computers to see

Isak Engdahl holds his thesis in fron of the Lux building.

Artificial intelligence powers everything from facial recognition to self-driving cars, but what does it really take to teach a computer to “see”? Isak Engdahl, who recently defended his doctoral thesis at Lund University, takes us inside the labs where computer vision systems are built, and reveals the human work behind the technology.

Black and white portrait of Isak Engdahl.
Isak Engdahl

On Friday 28 November, Isak Engdahl successfully defended his doctoral thesis in sociology at the Lux Auditorium. The research offers a new way of looking at the interaction between developers and AI. To understand the interaction between humans and machines, he studied the work on site at laboratories during fieldwork where he participated in meetings and interviewed staff.

What inspired you to study this topic?

“My research is driven by the need to analyse computational technology and understand how it shapes contemporary society. My doctoral work sits within the Show and Tell project, which investigates visualisation technologies across different environments.”

Isak Engdahl focused on computer vision, the field where many of these technologies originate.

“This gave me an entry point into studying contemporary AI through the laboratory activities that produce these systems. For me, understanding the societal impacts of these systems begins with understanding how they are made.”

What was the most interesting or surprising thing you discovered during your research?

“One thing that stands out is AI systems being described as black-boxed or opaque. I don’t deny that opacity is real, but my fieldwork shifted how I think about the problem. I saw researchers hard at work probing, interpreting, and negotiating with their models, sometimes making new inroads into understanding them.”

Like other scientific fields, “total knowledge” is rarely achievable. The practical question becomes how researchers handle uncertainty and organise the circulation of results.

“The limits of how knowledge travels offer a good example. Papers and conferences circulate some insights, but only in a narrow sense.”

Enormous resources

Much of the practical work shaping model development stays within specific groups and doesn’t move cleanly through formal channels. This raises questions about how much of the “state of the art” depends on knowledge that remains local and unevenly shared.

“I was struck by the engineering ingenuity embedded in daily lab practices and by the magnitude of industrial mobilisation. Billions in funding and infrastructure have gone into building AI. It makes me wonder whether strategies used to ‘scale’ AI could be adapted to tackle climate change.”

What do you want to explore next?

“I plan to investigate the ‘post-scaling’ era of AI as the initial rush to scale models settles. I’m also interested in human-AI interaction, how alignment dynamics emerge when people use generative systems, and how AI-assisted discovery changes scientific practice.”

Isak Engdahl’s work reminds us that AI is not magic, it’s the result of collaboration, negotiation, and creativity. By uncovering the social and technical processes behind computer vision, his research offers a deeper understanding of how these systems shape, and are shaped by, society.


What is computer vision?

Computer vision is the branch of AI that teaches computers to interpret and understand images and video. It is used in technologies like facial recognition, self-driving cars, and medical imaging.


The front cover of Isak Engdahl's thesis.

About the thesis

Isak Engdahl has written a thesis entitled Pixels and Weights: The Situated Work of Teaching Computers to See. Supervisors were Alison Gerber and David Wästerfors.

Read more about the thesis in Lund University’s research portal