Software Should Have Corners

This morning, somewhere between Tallinn and Helsinki, I took three photos of a ferry lounge because I could not stop looking at the chairs.

The room was full, but not in the way an airport gate is full. People had distributed themselves across it with surprising precision. A person alone had found a window seat. Families had taken the larger tables. Two people sat opposite each other where conversation was expected. Some passengers disappeared into high-backed booths. Others sat close to strangers without ever becoming socially involved with them.

Nobody had been given instructions for any of this.

There was no onboarding screen asking what kind of interaction I wanted today. No setting for social visibility. No toggle between public and private. No menu offering “alone, but not isolated” or “available for conversation, but only a little.”

The furniture did it.

I started thinking that the ferry lounge had more sophisticated privacy settings than most software.

A chair facing another chair creates one possibility. Two chairs angled slightly away create another. A sofa along the edge of the room allows several people to occupy the same zone while still pretending to be separate. A high-backed seat creates a tiny room without a door. A table gives people somewhere to place objects, but it also creates a temporary border. Put a backpack on the chair beside you and the border expands.

None of these things is absolute. That is exactly what makes them useful.

Physical space is full of gradients.

I can be alone. I can be alone while seeing other people. I can sit near someone without acknowledging them. I can make myself approachable. I can face another person directly. I can join a group. I can retreat without leaving the room.

Software tends to flatten these states.

Public or private. Online or offline. Following or not following. Friend or stranger. Member or non-member. Visible or hidden.

We build binary controls around relationships that are not binary at all.

A good seating area understands that people often want contradictory things at the same time. I may want to be around people without talking to them. I may want to watch what is happening while protecting myself from becoming part of it. I may want the possibility of interaction without the obligation of interaction.

The window seat is a beautiful example. It gives me something to look at that is not another person. I can sit in a social environment while orienting my body toward the sea. I am present without performing presence.

Software rarely gives me the equivalent of a window.

It gives me the feed.

The feed is a room where every chair faces everybody else.

That might be why so much digital social space feels exhausting. It has very little geometry. Everyone is exposed to the same central surface. Everyone is potentially addressable. Every post can become an interaction. There are few corners to occupy, few thresholds to linger in, few places where I can be visible without becoming available.

I think software needs more corners.

Not decorative corners. Social ones.

A corner is a place where fewer things can happen to me at once. It reduces the number of directions from which attention can arrive. A wall behind me means I do not need to monitor what is happening there. A high-backed chair performs the same function at a smaller scale. It changes the probability of interruption without banning interruption.

That is different from a privacy setting.

A privacy setting is usually a rule. A corner is a condition.

The difference matters.

Good interaction design is often described as making the right action obvious. But a room suggests another possibility. Maybe the designer does not always need to make one action obvious. Maybe the designer can create a landscape in which several good actions become possible and people can choose among them with very little explanation.

The ferry lounge does not force me into a flow.

It gives me territories, paths, edges, views, distances, surfaces and degrees of enclosure. I compose the experience myself.

This becomes even clearer when I look at the walkway cutting through the seating area. The wooden path is not really part of the lounge, but the lounge would fail without it. It lets people cross the room without constantly negotiating with people who are already seated.

A bad room makes you say “sorry” every ten seconds.

A bad app does something similar.

The interruption may be a modal window, a permission request, a forced tutorial, a confirmation screen, a notification, a menu that covers the thing I was using, or a navigation structure that makes me repeatedly back out of places I never meant to enter.

We tend to design screens as destinations and then treat movement between them as plumbing.

Physical space makes the opposite lesson unavoidable. Circulation is part of the experience. The route to the seat matters. The distance between the table and the walkway matters. Whether I have to squeeze past someone matters. Whether I can see where I am going before I stand up matters.

The path is an interface too.

Then there is territory.

A public chair becomes temporarily mine when I sit in it. A table becomes more mine when I put a coffee on it. A second chair becomes ambiguous when I put a bag there. The furniture has not changed ownership in any legal sense, but everyone nearby can read the state.

This is a very sophisticated interface behavior.

Objects modify permissions.

The bag says: probably occupied.

The half-finished coffee says: someone may be coming back.

A coat over the back of a chair says: this position has a history.

The system does not need a database field called reservedBy. The state is visible in the environment.

Digital products often hide this kind of state inside the machine. Someone is editing the document, but I only discover it after a conflict. Someone has claimed the task, but the interface makes the ownership subtle. A draft exists, but it is hidden in a different section. Something is temporarily mine, but nothing in the environment feels different.

Physical territory suggests that persistent state should sometimes have a spatial consequence.

If I have claimed something, maybe the product should look claimed. If I have left something unfinished, perhaps it should remain somewhere I can see it. If another person is present, maybe their presence should change the shape of the space rather than merely add a small avatar to the corner of the screen.

The most interesting thing, though, is what happens when people ignore the designer's intended use.

A four-person table becomes a one-person office. A chair becomes a luggage rack. Two tables get pulled together. Someone turns a seat toward the window. A family colonizes a whole cluster. A child treats the gap between the seats as part of the furniture.

The system survives appropriation.

In fact, appropriation is part of how it works.

Software is often much less generous. Every object has a type. Every action has an expected place. Every workflow is represented as a sequence. We talk about flexibility, but many digital products become confused the moment a person uses something for a purpose the designer did not predict.

A room cannot predict everything either. It solves the problem differently. It provides enough structure to be legible and enough looseness to be inhabited.

That may be one of the most useful interaction-design lessons in the whole lounge.

Do not design every action.

Design the conditions around action.

The difference also changes how I think about usability testing.

If I wanted to understand whether this lounge worked, I would not only ask passengers whether the chairs were comfortable. I would watch which seats filled first. Which seats remained empty even when the room became crowded? How close were strangers willing to sit? Which tables were effectively occupied by one person? Where did people put bags? Did anyone move furniture? Which routes became blocked? Where did children go that adults did not? Which seats were abandoned as soon as a better one became available?

The empty chair may tell me more than the occupied one.

The designer's plan tells me what the furniture was supposed to do. Occupancy tells me what people think it is for.

That is true of software too.

A feature list tells me what the team built. Analytics can tell me what people touched. But the more interesting evidence may be the workaround, the abandoned field, the screenshot someone keeps because the product has no memory, the spreadsheet beside the expensive enterprise system, the chat message used as a database, the browser tab left open for three weeks because there is no better way to preserve a state.

Those are the digital equivalents of putting a bag on the next chair.

They are not noise around the product. They are people furnishing it.

I used to think interaction design was mainly about the moment when a person does something and a system responds.

The ferry lounge makes that definition feel too small.

Interaction design can happen before anyone acts. It can be embedded in distance, orientation, visibility, enclosure, friction and the number of possibilities available from one position. The design changes what feels natural before it changes what is technically possible.

A chair does not tell me to talk to someone. It makes talking more or less likely.

A wall does not tell me to relax. It reduces one direction I need to monitor.

A table does not tell other people to stay away. It gives my temporary territory a visible edge.

A walkway does not tell me where I must go. It makes one route easier than the alternatives.

This is interaction design without instructions.

Maybe software became too obsessed with commands because commands are easy to represent. Buttons, menus, forms and toggles are explicit. They translate neatly into requirements and event tracking. We can point to the moment when the user clicked the thing.

Conditions are harder to measure.

How approachable did the interface feel? How exposed did I feel while using it? Did I have somewhere to retreat without leaving? Could I remain nearby without committing? Did the product give me territory? Could I understand what belonged to me and what belonged to someone else? Was there an edge where I could stop and look around?

Rooms have been answering these questions for a very long time.

Software is still learning how to ask them.

Connected reading

• The City Is an Interface
• The Best Products Aren't Apps. They're Cities.
• A Good Building Can Still Make a Bad City
• The Product Happens Between the Screens

AI authorship note

This essay began with Kris Haamer photographing the seating areas of a ferry lounge between Tallinn and Helsinki and asking what they could teach him about interaction design. Kris identified the Raw Drafts angle as the contrast between the nuanced social geometry of physical space and the binary states of software. AI helped develop, structure, edit and implement the essay. Kris supplied the observation, photographs, framing, title direction and publication decision.