Course description
Images generated, classified and retrieved by machine-learning systems increasingly shape what researchers (and the public) see. This workshop pairs practical skills with media-critical thinking so researchers working with images and visual analysis can use AI responsibly. We begin with a non-technical introduction to CLIP (Contrastive Language–Image Pre-training): a machine learning model trained by OpenAI on image–text pairs that “reads” short text prompts, assigns meaning to them, and retrieves the images it “sees” as matching those meanings — enabling natural-language and image similarity-based search, as well as automatic labelling of large image collections. CLIP is powerful, but its outputs are not neutral—they reflect the training data, cultural assumptions and technical choices that produced them.
The first, humanities-focused half uses short, active exercises on a curated image dataset to practise interrogating what CLIP “sees”: noticing what is foregrounded or elided, tracing likely sources of bias and error, and making human, contextual readings of machine outputs. Throughout we question how delegating the act of looking to algorithms reshapes our analytic habits—how embeddings function as new “frames” of vision—and what literacies and ethical attentions researchers need when their primary visual interface is a model like CLIP. After a short break, the second half is a guided, practical session in a Google Colab notebook (no prior coding required; cells are prewritten). Participants may use the workshop dataset or upload their own collections (archives, social-media datasets, and/or personal photos) to run text queries, similarity searches and basic clustering; we’ll also discuss result interpretation, simple error analysis, and ethical considerations for research and public deployment. Throughout, we emphasise combining digital methods with visual-culture theory so your use of AI vision is ethically informed and interpretively careful.
Target audience
This workshop is designed for an interdisciplinary audience with an interest in visual research and responsible use of AI vision tools:
- Master’s students, PhD candidates, and researchers in the humanities and social sciences (film/media studies, art history, visual culture, communication, sociology, anthropology).
- Digital-humanities practitioners, librarians, archivists, museum professionals, and data curators working with image collections.
- Educators who want to integrate AI-based visual methods into teaching.
- Relevant fields: Film & Media Studies, Art History, Visual Culture, Digital Humanities, Library & Information Science, Archival Studies, Museum Studies, Communication, Sociology, Anthropology — and interdisciplinary projects that combine these fields with computational methods.
Learning outcomes
- Explain, in plain terms, what CLIP is and how it connects written prompts to images.
- Understand how machine-vision systems influence what we see and how that reshapes visual scholarship.
- Perform basic CLIP workflows (text queries, similarity search, simple clustering).
- Critically evaluate model outputs—identify likely errors, biases, and omissions—and interpret results alongside human contextual readings.
- Discuss ethical considerations and responsible practices for using CLIP in research and public projects.
Prerequisites
- Beginner level — no prior coding or machine learning experience required.
- Recommended: basic comfort with using a web browser, uploading/downloading files, and editing short text (for writing prompts).
- Useful but optional: familiarity with basic concepts in visual/media analysis or interest in image-based research.
Technical and Material Requirements
- Laptop (required). Bring a charged laptop with a modern browser.
- Reliable internet access (required). The hands-on Colab session runs in the browser and needs a stable connection.
- No software install needed. All code runs in the browser (Google Colab); cells are prewritten and annotated.
- Optional — your own image collection. If you want to test CLIP on your own material bring a folder of images or be ready to upload from your device. Recommended formats: JPG / PNG; per image <10 MB; total collection ideally ≤ 500 images / ≤ 500 MB to keep the notebook responsive.
- File naming & rights: Please ensure you have the right to use any images you bring and avoid uploading sensitive or personally identifying images unless you are certain it’s appropriate.
- We provide: A curated dataset for in-workshop exercises and a shared Colab notebook (link distributed at the workshop).
Lunch
- There will be a complementary buffet-lunch at 12:00-13:00 from the Sverdrups spiseri (Georg Sverdrups Hus, 2nd floor).
- After you get your lunch, join us at Store møterom where we have reserved seats for #DSDays2026 participants.
Number of participants
After the soft deadline, a waiting list will be automatically handled upon de-registrations.
Contact
If you have any questions regarding this course, or anything else about the DSDays 2026 event, please contact us per email: digitalscholarship@ub.uio.no