ACM RecSys 2026 Introduces a Rather Different Call: Tell the History of Recommender Systems in 60 Seconds

ACM RecSys 2026 has quietly added a rather unusual call for contributions. And quietly is the important word here: the call appeared only a few days ago. With submissions due on September 4, this is very much a short-notice opportunity. So, if you want to participate, do not put this one onto the usual academic “I will look at it next month” pile.

The new format is called the 20×20 Retrospective Track. It is part of the celebrations for the 20th ACM Conference on Recommender Systems, which will take place in Minneapolis from September 28 to October 2, 2026. RecSys started in Minneapolis in 2007, so returning there for its twentieth edition provides a suitable occasion to look backwards as well as forwards.

The format is unlike anything I remember seeing at RecSys. There is no paper to write. No four-page limit. No appendix. No reviewer asking for three additional datasets two weeks before the camera-ready deadline. Instead, participants are asked to submit a video of at most one minute.

The conference organizers have curated 20 milestones that they consider important for the development of the recommender-systems community (PDF see below). They span three categories: Foundations, Applications, and Community. Participants choose one milestone and record a short personal reflection. The question is not “What is the complete scientific history of this milestone?” but rather: Why did this particular development matter to you? What did it change in your work, your thinking, or your experience of the community?

The list itself is worth browsing. It includes obvious scientific and technological landmarks, but also events that helped create recommender systems as a community. The Netflix Prize is there. So is the first ACM RecSys conference, the RecSys Challenge, Women in RecSys, and even RecSys karaoke. User-centered evaluation frameworks such as ResQue are represented as well. This makes the exercise broader than a conventional list of influential papers. It acknowledges that a research field consists of algorithms, applications, institutions, people, and, apparently, occasionally microphones.

I like this format for RecSys in particular. Recommender-systems research has changed substantially during the lifetime of the conference. Matrix factorization, context-aware recommendation, deep learning, fairness, user experience, conversational recommendation, and now generative and agentic systems have each changed what the community considers an important research problem. The current RecSys 2026 call itself reflects that breadth, including topics ranging from human-centered recommendation and societal impact to generative, agentic, and reasoning-based recommendation.

There is also something fitting about asking for personal recollections rather than another retrospective paper. Scientific histories tend to become clean in hindsight. One method leads to another, one benchmark replaces another, and a few highly cited papers become the official milestones. Actual research careers are less orderly. A competition may have brought someone into the field. A dataset may have enabled a PhD thesis. A user study may have convinced an algorithm researcher that optimizing RMSE was not the entire problem after all. Capturing those stories can tell us something that citation counts cannot.

Of course, the format also has limitations. A curated set of 20 milestones is necessarily selective. Any list that compresses two decades of recommender-systems research into 20 entries will omit developments that some of us consider central. Personal memories are also not the same as historical evidence. But that is arguably the point. The organizers explicitly ask for personal reflection rather than a comprehensive history, and selection will be editorial rather than based on formal peer review.

Selected videos will be shown during the main conference and may later become part of a dedicated playlist on the ACM RecSys YouTube channel. The conference program already reserves a 90-minute “RecSys 20th Anniversary Presentation” on Wednesday, September 30, so the retrospective will receive a visible place in the main program.

The only real problem is timing. This call has arrived very late. At the time of writing on August 17, the September 4 submission deadline is less than three weeks away, with notifications following on September 11. Fortunately, producing a one-minute video should require somewhat less effort than producing an eight-page RecSys paper. At least in theory. Anyone who has ever tried to explain twenty years of research in sixty seconds may disagree.

Still, this is one of the more interesting anniversary ideas I have seen at an academic conference. RecSys has spent twenty years studying how to select a small number of relevant items from a much larger set. Now the community is being asked to perform the same task on its own history: choose one milestone out of twenty, and explain why it mattered. That seems appropriately recursive for a recommender-systems conference.