The Memory Gap

YYou walk into a therapist’s office. You sit down. You start talking.

For the first few sessions, you piece your life together in response to some version of: “So, tell me about yourself.” You narrate your childhood. Describe your relationships. Explain the anxieties that brought you here. You try to give a stranger enough context to understand who you are, how you got here, and what might be wrong.

But you’re not giving them your life. You’re giving them what you can remember about your life, from where you happen to be standing today. And that distinction matters.

Therapy often begins with reconstruction because it has to. A therapist wasn’t there for the argument six months ago, the friendship that slowly fell apart, the year you were miserable at work, or the childhood moments you still think about twenty years later. The primary source is you. And human beings are remarkably bad at accurately reconstructing their own past.

The Retention Problem

Allison Harvey, a psychology professor at UC Berkeley, encountered the problem from the other side. She enrolled in an 11-week parenting course, took extensive notes, completed every homework assignment, and engaged intensely with the instructors. By any reasonable measure, she was an ideal student.

When the course ended, she realized she had forgotten much of what she had learned. Her frustration became a research question, and what she found was sobering: research on treatment memory suggests patients can retain surprisingly little of the information discussed during clinical encounters.

“I don’t think us humans realize how fallible our memories are,” Harvey told UC Berkeley. “We’ve got decades of psychological research showing that we fail to encode memories accurately. We fail to store them accurately. And we fail to retrieve them accurately.”

Memory is not an archive. We experience something, encode only part of it, retain only part of that, and later reconstruct what remains. Then we use that reconstruction to explain ourselves.

The Transmission Problem

For therapy, forgetting is only half the problem. The other half is transmission.

Imagine trying to compress a relationship into 50 minutes. Which fights do you mention? Which do you forget? What happened first? How did you feel at the time, and how much has the way you feel now changed the way you remember it? Was a particular behavior typical, or does it stand out precisely because it wasn’t?

You make hundreds of editorial decisions without realizing you’re making them. Then language compresses the experience again. “I felt ignored.” “We were having problems.” “My childhood was pretty normal.” Each sentence can represent months or years of events, conversations, contradictions, and changing interpretations.

By the time an experience reaches another person, it has passed through several layers: what happened, what you noticed, what you remembered, what you can retrieve now, what you choose to say, and finally what the other person understands.

None of this means your account is false. It means it is incomplete. And for many of the conversations that matter most, incomplete recollection is the only record in the room.

The Other Memory Problem

Your therapist has one too.

Consider Mary, a 31-year-old who spent 18 months in therapy. Her therapist repeatedly confused important details of her life: asking about a brother she didn’t have, confusing Pilates with yoga, and recalling relationship and reproductive details incorrectly.

“All these mistakes have eroded my sense of confidence and trust,” Mary said. “It makes me think, you’re mixing me up with somebody else, and are you potentially discussing my issues with that person?”

The point isn’t that therapists are uniquely forgetful. They’re human too. One person is reconstructing a life from memory while another is trying to understand it, contextualize it, and retain the relevant details across dozens of conversations and many different patients.

Therapy already has tools designed to help with this: clinical notes, assessments, journals, homework, and treatment plans. Good therapists check assumptions and revisit earlier accounts. But the underlying information problem remains. Most of the life being discussed happened somewhere else, and neither person in the room has a particularly complete record of it.

What If You Brought Some of It With You?

What if, instead of relying almost entirely on recollection, you could bring some of the original context with you?

Not a comprehensive recording of your life, and not a machine-generated psychological profile. Just the pieces that already exist: the text exchange you keep thinking about, the voice memo you recorded ten minutes after the fight, photographs from a period you now remember differently, journal entries written before you knew how the story would end, the playlist you played obsessively during one particular summer, or a note you wrote to yourself three years ago and forgot existed.

Individually, these artifacts prove very little. Together, they can provide something memory alone cannot: contemporaneous context. They are not hard evidence in the courtroom sense. They are soft evidence—traces left behind by a life while it was actually being lived.

The Trace

Circa8 is developing a way to organize that material into something called a trace: a curated body of personal evidence assembled around a particular question, period, relationship, or conversation.

The distinction between curation and collection is important. A trace is not supposed to be a system quietly recording everything you do, nor is it an AI constructing a psychological profile behind your back. You choose the material. The system helps you organize it.

A trace might contain photographs, journal entries, messages, voice notes, videos, documents, or recordings. AI can make that material searchable, establish chronology, surface recurring subjects, and identify connections that would otherwise be difficult to see across hundreds of artifacts. What it should not do is decide what those connections mean.

That boundary is deliberate. The purpose isn’t to build an AI that understands you better than you understand yourself. It is to give you—and the people you choose to involve—better material with which to understand your life.

The Evidence

There is already evidence that information generated outside the clinical encounter can add useful context to mental-health care. A 2017 study published in the Journal of Medical Internet Research, for example, examined whether behavioral signals derived from smartphones—including communication and mobility patterns—could help predict clinically assessed symptoms of depression and PTSD.

More recent research has explored a different question: even if personal digital information could be useful, would people actually want it incorporated into mental-health care? The 2024 Tracing Tomorrow project examined young people’s attitudes toward personal sensing for mental health and found that willingness depended heavily on whether the use of that information felt trustworthy, relevant, and acceptable.

That points toward a model very different from passive surveillance. Rather than collecting everything and asking an algorithm to diagnose a person, the individual decides what becomes part of the record, why it matters, and who gets to see it. The technology provides organization; the human retains agency over both the evidence and its interpretation.

The Mechanism

Imagine that before a therapy session you want to understand a difficult period in a relationship. You gather a screenshot of a conversation that bothered you, a voice memo recorded immediately afterward, three journal entries from different points in the relationship, a photograph from a day you remember as happy, and a message you sent a friend that same night saying you wanted to leave.

The system doesn’t look at those artifacts and announce that you were unhappy. It recognizes that they concern the same period and puts them into context. Now you and your therapist can examine them together and ask more interesting questions: Why do I remember that day so differently? Was this an isolated incident or part of a pattern? When did my language begin to change? What did I believe then that I no longer believe now?

The artifacts don’t answer those questions. They make the questions possible. Instead of spending the entire conversation trying to reconstruct what happened, you have something closer to the original material from which to examine it.

The Shift

That changes the role technology plays in the conversation.

Allison Harvey has argued for simple interventions that help patients remember and use what happens in treatment. “We can come up with augmentations that could be applied before a session, in the middle of a session or at the end of a session that are not costly,” she said. “They don’t have side effects, but they’re effective at improving outcomes.”

A trace extends that logic in the other direction. Rather than only helping someone remember what happened inside treatment, it can help bring relevant context from outside treatment into the room.

The aim is not perfect recall. It is a richer starting point.

Beyond Therapy

Therapy makes the memory problem unusually visible, but it is hardly unique to therapy. Performance reviews compress twelve months of work into a meeting largely governed by what an employee and manager happen to remember. A trace could reconstruct projects, decisions, accomplishments, setbacks, and feedback while they were actually happening.

The same idea applies to difficult conversations with family. Two people can remember the same decade differently without either deliberately lying. Old photographs, letters, recordings, and messages cannot determine whose interpretation is correct, but they can provide context that neither person can reliably recreate years later.

Relationships present an even more delicate version of the problem. Partners routinely disagree not only about what an event meant but about what happened at all. Contemporary artifacts should not become ammunition or a mechanism for proving one person right. Used carefully, however, they can reveal something more interesting: how differently two people experienced the same period.

Across all three cases, the principle is the same. Technology does not need to replace human interpretation to be useful. Sometimes its most valuable role is simply to give human interpretation better source material.

The Memory Workaround

We ask memory to perform a job it was never designed to do. We expect it to preserve decades, maintain chronology, retain context, and then retrieve the relevant pieces on command during some of the most consequential conversations of our lives.

Memory does something more complicated. It selects, compresses, updates, and reconstructs. The person remembering an experience years later is not quite the same person who originally lived it, which may be precisely why encountering an old artifact can feel so strange.

You find something written in your own voice that you no longer remember writing. A photograph complicates the story you have told yourself about a particular year. A message reveals that you knew something months before you remember realizing it. The artifact does not necessarily tell you which version is true. It simply gives you another piece of the record.

A trace does the same thing at a larger scale. It doesn't replace memory, and it doesn't promise an objective account of a life. It gives memory something external to work alongside: material created closer to the moment, before years of hindsight had the opportunity to reshape it.

That may be enough to change the quality of the conversation. The breakthrough isn't replacing the therapist, the conversation, or even memory itself. It's giving people a better record from which to ask better questions.