why we do this
This is a point of view, not a verdict. Everything below is our reading of published research on how people live together, a synthesis we use to decide what roomies should and shouldn't do. We name our sources and their limits, we never invent a statistic, and we try hard not to dress up correlation as cause. Where a finding comes from a different setting (couples, lab studies, common-pool resources like fisheries), we say so plainly, because borrowing a result is an argument, not a proof.
We borrowed the shape of this page from reports like Hinge's “Love & ADHD”: the move where you stop treating friction as a personality flaw and start treating it as a mismatch in habits, schedules, and unspoken expectations. That reframe is the whole thesis: most roommate trouble isn't a bad person, it's an unaligned household. Here is what follows from that.
1. Match on habits, not personality
The intuitive idea, “find someone like me and we'll get along,” is mostly wrong about personality. The claim that global personality similarity predicts roommate satisfaction does not hold up; the better-supported pattern is that satisfaction tracks perceived similarity in concrete living habits and communication patterns, not personality type or shared taste (Niu & Brown 2023; Martin & Anderson 1995; and earlier, Carli, Ganley & Pierce-Otay 1991). Note the word perceived: what predicts satisfaction is how similar the habits feel, which is why a good first meeting can be a trap: it reads rapport, not routines. So roomies screens and surfaces the things that actually vary day to day (sleep schedule, cleanliness, noise, guests, how you handle money) rather than sorting people by “type.”
2. Treat the household as the unit, not the individual
A shared home is an interdependent system, not two people who happen to share a wall, so the household, not the individual, is the thing worth designing around (a systems framing; Erb et al. 2014). Interdependence here is not a metaphor: in a study of 103 randomly assigned first-year roommate pairs, students who drew a roommate high in a depression-linked thinking style were more likely to take on that style themselves over the following months (Haeffel & Hames 2014, a genuine causal design, because the pairing was random). We read that as: who you live with shapes you, for better and worse, so the match matters more than “just find a room.”
3. Surface fairness: don't just split evenly
An even split and a fair-feeling split are not the same thing, and the gap between them is where a lot of money fights live. How fair a split feels (perceived fairness) is associated with how a shared arrangement holds up, though the actual numbers carry weight too; we do not claim “perceived fairness beats the real split, full stop.” The cleanest evidence that explicit, agreed expectations lower relational uncertainty comes from studies of romantic couples (Le & Aune 2011), so we treat it as a labeled analogy, not a roommate finding. roomies' job here is to make the split visible and negotiable (who pays what, and why), not to declare a single “correct” number.
4. Make expectations explicit, early
Most of what people fight about (quiet hours, guests, the dishes, the thermostat) was never actually agreed, just assumed. Naming expectations early reduces the uncertainty that friction grows in. The mechanism we lean on is implementation intentions: specific “when X, I'll do Y” plans, which across many domains have a modest-to-moderate average effect on whether plans actually get carried out (meta-analytic d ≈ 0.14–0.65; Gollwitzer & Sheeran 2006). You'll sometimes see a dramatic single-study figure (one classic study moved a behavior from roughly 22% to 62%); that is the original study only, not a general rate. And an important hedge we won't drop: this mechanism has been tested on exercise, dieting, and voting, not specifically on roommate agreements. We think it transfers; we can't prove it does yet. So roomies prompts the explicit conversation early, before move-in.
5. Design the commons
A shared apartment is a small commons: shared space, shared bills, shared chores, no boss. Elinor Ostrom's work identified eight design principles that help groups govern shared resources well (clear boundaries, costs and benefits roughly proportional to use, low-cost ways to resolve conflict, and so on), and a later paper argues these generalize beyond fisheries and forests to almost any group that has to cooperate (Wilson, Ostrom & Cox 2013). To be exact: this is research supporting the use of commons-governance principles in shared living. Ostrom did not study roommates. We find the framing genuinely useful, good households run on lightweight rules and a way to fix problems, not on goodwill alone, which is why roomies leans toward written agreements and clear chore systems over “we'll figure it out.”
6. Name the behavior, not the person
The same dirty pan reads as “they're inconsiderate” or “they had a brutal week” depending on the story you tell about why. Roommate conflict is driven heavily by these attributions (the perceived intent behind an act) more than by the objective disagreement itself, and integrative, problem-naming communication does better than avoidant or blame-trading styles (Sillars 1980). Your own communication is the biggest lever you actually control. So the language we use, and the language we encourage, targets the behavior and the fix (“can we set quiet hours after 11?”) rather than the character of the person.
7. Protect privacy and autonomy as much as sociability
A good roommate match is not maximum togetherness. People need to dial closeness up and down (solitude, a closed door, control over their own space), and a home that only optimizes for “more social” gets the balance wrong. We hold this as a design value more than a settled empirical result: this principle is asserted from theory and is pending a firmer citation in our evidence base. Practically, it's why roomies treats privacy and personal autonomy as first-class (in what we ask, what we surface, and what we leave alone) rather than nudging everyone toward more contact.
8. A good match is a modest wellbeing intervention
Who you live with touches how your year goes: your stress, your sleep, your sense of having somewhere that feels like home. At the population level, social connection is one of the more robust correlates of mental and physical health (Holt-Lunstad 2024). We state this carefully, with the strongest possible hedge: those are population-level associations, and we make no claim that using an app changes anyone's mortality, GPA, or mental health. A better-aligned roommate is a small, plausible improvement to daily life, a modest intervention, and we'd rather under-claim that honestly than oversell it.
9. Publish honestly
The last principle is about this hub itself. Every number we publish traces to a named public source or a clearly-labeled first-party observation; we don't invent proprietary statistics, and at our current scale we have no honest generalizable data of our own to report, so we publish literature synthesis and label any first-party note as exactly that. Vendor figures are attributed as vendor. Correlational findings say “associated with” or “predicts,” never “causes.” And every study or data piece ends with its sources and limits, including this one.
what this means for the app
Put together, these principles point one way: roomies' job is to screen and surface the habits, schedules, and expectations that actually predict living well together, and to make agreements and fair splits easy to set early. It is not a compatibility guarantee, and it doesn't manufacture trust. Those aren't things software can promise. It's a tool for having the right conversations before you sign, grounded in the best reading of the evidence we can offer, limits and all.