Why do I forget to text people back?
Because the intention arrives at a moment when you cannot act on it, and intentions decay. The best-evidenced fix is not willpower but specificity: deciding in advance when and where you will do a thing has a medium-to-large effect on whether you do it — d = 0.65 across 94 tests and more than 8,000 people. A reminder that says "message Sara" is a nag. One that says "on the bus home tonight, ask Sara how her mum's surgery went" is an instruction you can follow.
Gollwitzer & Sheeran (2006), implementation intentions
The full answer →
How do I remember details about my friends?
Write the detail down within a few minutes of hearing it, and make sure something brings it back before the next conversation rather than leaving it in an archive. The lifelogging research is unambiguous on the second half: capturing everything produced enormous archives and no useful applications, while reviewing a small selective capture reliably cued recall months later. Capture is cheap; the value is made at review.
Sellen & Whittaker (2010), Beyond Total Capture
How to actually do it →
Is it weird to message someone after a long silence?
No, and the gap between what you expect and what actually happens has been measured repeatedly. Recipients of an out-of-the-blue message appreciate it more than the sender predicts, and the effect gets larger the more surprising the contact is. People also over-predict how awkward a deeper conversation will be and under-predict the connection it produces. The reluctance is a prediction error, not politeness.
Liu, Rim, Min & Min (2023), The Surprise of Reaching Out
What to send instead of an apology →
Why do friendships fade when nothing went wrong?
Because the shared context disappears and nobody initiates. In a Dutch panel study that re-interviewed people seven years apart, only 48% of a personal network was still in it — and the network had not shrunk, the names had changed. The common cause of a lapsed tie was a lack of opportunities to meet, not conflict. Roberts and Dunbar found the asymmetry that follows from this: over eighteen months, emotional closeness to friends declined once regular contact stopped, while closeness to family survived the same gaps. Friendship is the relationship that carries an ongoing maintenance cost.
Mollenhorst, Völker & Flap, Social Networks
Friendships don’t end, they drift →
Is a personal CRM a good way to keep up with friends?
The category imports machinery that close relationships do not tolerate. Tracking inputs and balances — who owes whom a message, how long since last contact — is characteristic of exchange relationships, and applying it where closeness is wanted damages the thing it is meant to help. Separately, measuring an enjoyable activity increases how much of it people do and decreases how much they enjoy it, with the drop persisting after the measurement stops. Track what someone needs and what they told you; do not score the relationship.
Etkin (2016), The Hidden Cost of Personal Quantification
Why we’ll never ship a streak →
Should an app write the message for me?
No. Messages known to involve AI are judged less trustworthy, less authentic and less useful, and the penalty grows with the degree of AI involvement. There is also a disclosure paradox: people say disclosure matters and then punish it when it happens. An assistant that drafts your outbound text manufactures exactly the artefact that is losing its value. Kept never writes your message — its AI faces inward, transcribing and organising what you said.
Hancock, Naaman & Levy, AI-mediated communication
The research on the AI penalty →
Does Kept listen in the background?
No. The microphone opens when you hold the button and closes when you let go; there is no other mode. Around a fifth of people do not realise their voice devices are always on, and privacy concern rises sharply once they find out — so this is a design constraint rather than a setting. Kept also records what you noticed, never a recording of the other person, because it holds notes about people who never agreed to anything.
Malkin et al., privacy controls for always-listening devices
Why push-to-talk is architectural →
Does Kept reduce loneliness?
There is no evidence that it does, and we do not claim it. Kept has never been tested in a trial, and no trial of it is planned. The research on this page describes how relationships work — how they lapse, what reaching out actually does, why remembered detail matters. It does not describe what Kept does to them, and we are not going to imply otherwise.
Every study, with its limits →
