The research behind it
What "grounded in research" means here
The phrase gets used loosely, so here is what we mean by it.
What it means: the behaviour we designed is drawn from peer-reviewed literature. Named theories, by named researchers, most with meta-analyses behind them. We can point to a source for each. Where a famous finding failed to replicate, we treated it as failed.
What the app does, and why
Each of these is something you can watch the app actually do. The names are enough to find the original research.
It waits until you ask
The app doesn't volunteer advice you didn't request. Motivational Interviewing (Miller & Rollnick) found that help offered before someone is ready measurably increases resistance, the opposite of the intended effect. Scaffolding research (Vygotsky; Wood, Bruner & Ross) adds the other half: effective help is calibrated to where the person actually is, and then withdrawn. Self-Determination Theory (Deci & Ryan) explains why autonomy matters so much to whether a change sticks.
It suggests; you decide
Nothing is sent on your behalf. Every draft is yours to accept, edit, or throw away. Dietvorst's work on algorithm aversion found that people will accept an imperfect system far more readily when they keep control over its output, and abandon a better one that takes control away. Research on automation bias points the same direction: show the person the material, don't just hand down a conclusion.
It encourages a pause before you send
Loewenstein's hot-cold empathy gap describes something most people recognise: the calm version of you cannot accurately predict what the angry version wants, and vice versa. They are working from different preferences. Lerner's work on anger and decision-making shows that carryover affects choices well after the triggering event. A short delay is one of the better-evidenced interventions available.
It treats a rough moment as repairable
Tronick's still-face research, among the most replicated findings in developmental psychology, showed that healthy relationships are not the ones without rupture. Attunement is intermittent; what predicts security is the reliability of repair afterwards. John and Julie Gottman on repair attempts, and Fincham & Bradbury on how the story you tell about someone's behaviour shapes what happens next, fill in the rest.
It asks the people who know you, separately
The Mirror collects reflections independently rather than in a group. That design is deliberate: work on the independence requirement in collective judgment (Galton; Lorenz and colleagues) shows that once people can see each other's answers, agreement stops being evidence. It converges without getting more accurate. Edmondson on psychological safety covers the other risk: honest input only arrives when it's safe to give.
It asks for a specific next step
Reflection ends with something concrete rather than a resolution. Implementation intentions (Gollwitzer & Sheeran) is one of the most robust effects in behavioral science: if-then plans roughly double goal attainment compared with intention alone, and the specificity is what carries the effect. Paired with Kristin Neff on self-compassion held together with accountability, since self-criticism predicts avoidance rather than change.
Popular findings we deliberately don't build on
A page like this is worth little if it only lists the wins. Several well-known ideas did not survive scrutiny, and we avoid leaning on them even though they would make for easier marketing:
- Willpower as a fuel tank. "Ego depletion", the idea that self-control drains like a battery, largely collapsed under large-scale replication. We don't model your capacity that way.
- "Just don't think about it." The illusory truth effect (Hasher, Goldstein & Toppino) shows repetition raises believability on its own. Unchecked claims don't quietly fade, so we don't assume they will.
- Gottman's magic ratio. The widely quoted positive-to-negative numbers are directional, not thresholds. We don't present any ratio as a pass mark for a relationship.
- Growth as a guaranteed result of hardship. Post-traumatic growth is genuinely contested, and a good deal of it may be retrospective storytelling. We never promise that a hard thing will make you better.
- Confidence as a credibility signal. Smooth, fluent delivery is a poor detector of who is telling you the truth, so we don't treat it as one.
This is not therapy
Questions about a specific claim
If you'd like the citation behind any statement on this page, ask via the contact form and we'll send it. If you think we've got something wrong, we especially want to hear it, being corrected is the entire point of the thing we built.