The Model of SQUARE Research Retreats

Three colleagues have recently asked me about how we organize the SQUARE retreats. None of this is revolutionary, but when three people ask independently, it may make sense to write this up. Let’s reframe a research retreats from dense seminars of talks to mostly quiet work with just enough structure to keep people accountable and effective.

We held six of these retreats over the past years, starting in Helenekilde in 2015, then Dragør (‘22), Świnoujście (‘23, ‘26), and Sandbjerg Gods (‘24 and ‘25). There is a tentative plan to go to Eckebacken in Southern Sweden in 2027. There is a photo from each of these retreats somewhere in this post. The post captures “the way we do it”, for folks that are seeking to design their own retreats.

Why?

The primary goal of our retreats is to accelerate research for the entire research group, both juniors and seniors. For juniors (PhD students, assistants, postdocs) it is meant to be a motivating event, when a spell of dedicated time with accountability helps to deliver work, and break writing blocks. For seniors (faculty) it is really a retreat, an escape from other duties to create space reserved for research.

Another goal is to understand who is working on what in the group (an update on research activities). This goal is secondary. We are not interested in a comprehensive mini-conference, in which most people doze off most of the time.

Specifically, it is not a goal to increase within-group interaction, joint collaboration, schedule many research supervision meetings, etc. It is not a trip that is devoted to achieving some strategic goal set by the upper management or the head of group. I strongly believe that strategic initiatives are less helpful than diligent research. Strategic ideas should be developed in other spaces, possibly by writing grant proposals, and not by getting the attention of the whole group, which in aggregation is very expensive, and not equally useful to people at all career stages. So no top-down goals, but progress according to a bottom-up constructed plan.

This does not mean that research group bonding and interaction does not happen. It does, but mostly thanks to secondary means: common meals, sharing entertainment time, walk-and-talk plan and short research presentations (details below).

How? The Format

The key is to create dedicated time for work, with lightweight social pressure and easy access to supervisors and other group members. This is what creates the desired acceleration of research work.

We do not include Dagstuhl-style long brainstorming breakouts in the program. They sometimes happen ad hoc, when someone’s work gets stuck or when a discussion naturally emerges, but they are not the main format. Some colleagues were surprised that we travel to spend time together and then use most of the time not talking to each other. Well, what can I say—we just have a different purpose: getting things done. The conversations happen around the work, in walks, meals, coffee breaks, and late-evening discussions, not replacing work.

For Whom?

We want everybody who is employed in the group to join. Broad participation is important for the main goals: impact on research productivity and team building. All PhD students, postdocs, and faculty in the group are invited. We also include some part-time employees, for instance student programmers, if they are active in the period, or if we want to lure them into a PhD studentship. We also include external members who are not employed at ITU but have an affiliation with us: industrial PhD students and postdocs, as well as their company supervisors. The externally affiliated group members spend time at ITU only occasionally, so this is one time a year when the broad team can meet in one place. I find it very rewarding.

Finally, we always add 1–2 external guests, usually collaborators in current research projects, papers, or grant proposals. This brings external inspiration in both directions. Typically, the guests are somewhat senior, which helps to avoid the silo effect: the group of senior members of SQUARE changes slowly, so there is a risk of stagnation. Often, we invite guests so that the PhD students and postdocs who collaborate with them can accelerate their work during the week, thanks to easier and more frequent contact.

One issue is that this format does not scale to a very large group. Our last group of 20-ish participants is getting close to a maximum that still sustains a working attitude (as opposed to a noisy crowd). I hope we are not growing more, and the size of the group might affect the participation in the future.

When?

I have grown very fond of a full week (5 days) that does not encroach on the weekend. Arrival on Sunday evening, as at Dagstuhl, is fine if it cannot be avoided. Each year, several papers get created or seriously accelerated during the retreat. A short retreat works if the goal is intense interaction. But for a work retreat, things need time to unfold. Developing paper-worthy material takes time; a day or two is not enough. In the past, we had to shift the retreat to the weekend because of limited room availability. I learned then that motivating myself and others to work effectively during the weekend is much harder. We all just want to get home. Since then, we book early enough to get the weekdays.

My experience is that a retreat should not be scheduled during a relaxed period of the year, such as summer or winter break. These periods are already reserved for vacation and, if not, they are excellent research periods when we can work undisturbed. A retreat has a stronger effect during a relatively busy part of the academic year, when teaching-related and admin-related meetings disturb us a lot. A full meeting-free week (calendar block) refocuses us on research when this normally does not happen. The retreat is not meant to just move research to another location; it should add to the research time budget. At ITU, we have found that the three weeks right after lectures end and before the written exams start are optimal, which means mid-to-late May for us. Before that, people are too busy with lectures; after that, they are too busy with exams (both non-negotiable). A nice side effect is that we avoid the peak tourist season at most locations.

Where?

Ideally, the location should support immersion without making the logistics too painful. Biking distance, or ordinary local-transport distance from Copenhagen, is too close. It tempts people to go back home in the afternoon. I understand the excellent reasons for doing this, but it breaks immersion. I prefer to lose a little on participation, if that gives us better immersion.

We have sometimes gone a bit further, to Świnoujście, because the trip, even though longish, can itself become useful. The ferry gives us an extended work time. The group starts to bond already on the way, and people usually also work on their retreat tasks on the way there and back.

Some of us like remote, rural, natural locations, for instance Sandbjerg Gods, which gives a Dagstuhl-like sense of isolation. Several younger group members prefer small towns with convenience stores, small restaurants and cafés, and some limited night entertainment. Both kinds of locations have worked well for us so far. We avoid big cities.

I find it very useful to reuse the same location multiple times. In fact, my dream is to find an ideal location and just continue using it. So far this dream has only been partly achieved. Perhaps we have not found the ideal location yet, or perhaps some of us are too eager to try new things. Anyway, every year I start the discussion of where to go with the list of places we already know.

Finding a new location every year is a burden for organizers and participants. With a familiar location, there is no need to understand travel connections, coordinate trips from scratch, or renegotiate every detail with the hotel. The group becomes more autonomous and requires less coordination. There is also a lower risk of mishaps: too small a meeting room, too bad food, noise, low comfort, etc. Finally, the same location helps habitualize the mode of operation. Once we arrive at a familiar retreat location, our minds immediately get into retreat mode.

Retreat Outside the Retreat (Preparation)

Let us start with the timeline.

An example week looks roughly as follows:

A few other practical points matter as much as the timing.

What tasks work? The format is not optimized for vague exploration. It works best for finishing things, rewriting texts, programming concrete features, fixing bugs in research code, or reading a specific body of work. The key is to bring everything you need for the task: a laptop with a working build environment, a reference list, paper printouts, etc. Some scaffolding therefore needs to be done before you travel.

Finances: We piece together project money, department support, and accounts controlled by individual group members (pocket accounts, individual projects, etc.). Travel is paid by each participant from an account they control. This greatly simplifies the financial logistics. Otherwise the person handling the travel payment would have to deal with too many special cases. Pretty much every person has their own constraints on time, starting point, route, etc. We do not want to make anybody suffer learning that. Accommodation, board, and meeting rooms are paid centrally by the group from department and project funds.

Modeling cost: Sending about 20 people to travel together for a week is not cheap. Cost is always a key selection criterion for locations, both in time and money. To compare locations, I use a simple model. First, we estimate the typical per-person travel cost from Copenhagen, not using the cheapest possible tickets, since some people travel from more expensive locations, book late, etc. Then we compare hotel offers by calculating the average per-person daily cost for a given group size, including all costs charged by the hotel, for instance meeting rooms. The model is fairly noisy, but it helps estimate which locations are actually cheaper. Often the cheap locations per day are quite expensive per trip, so this has helped us make sound decisions in the past.

I am genuinely proud of these retreats. They are simple in design, but they have become one of the most important things we do as a group.

Credits. The design emerged from contributions and experiences from the RWTH Aachen MOVES group (they run fabulous retreats), Claus Brabrand, Christoph Seidl, Mahsa Varshosaz, Raul Pardo, Eduard Kamburjan, Peter Sestoft, Debasmita Lohar, Irina Shklovski and the HCI Section at the University of Copenhagen.