Research

Beyond intuition. Grounded in evidence.

MOSAIC’s lenses start from decades of published psychology and language research, and licensed clinicians review how we turn it into a report. Here is what that means, and where its limits are.

Why language

The words we choose carry patterns: who we center, how we apologize, what we avoid saying. Researchers have been measuring them for decades. MOSAIC reads them across months or years, not moments.

The evidence chain

From a study to a sentence in your report.

  1. 01

    It starts with published psychology

    Every lens is built on constructs researchers already study and argue about in the open: the Big Five, attachment, emotion regulation, narcissism, coercive control, and Gottman’s research on couples.

    How the lenses work
  2. 02

    Words carry measurable signals

    Decades of research show that pronouns, emotion words, absolutes and repair attempts relate to personality and emotional patterns. MOSAIC looks for those signals across months or years of messages.

    Read the studies
  3. 03

    Every finding shows its working

    Findings link back to the messages behind them and carry a confidence level. When the evidence is thin, the report says so rather than guessing.

    See a sample report
  4. 04

    Clinicians review how we say it

    Licensed psychologists, psychiatrists and counselors review our frameworks, the wording of findings and our safety language before it reaches you.

    What they review

Clinical review

Reviewed by people who do this work.

We work with licensed mental-health professionals across three disciplines. They don’t read your messages. They review how MOSAIC reasons, what it says, and how it says it.

Psychology

Clinical psychologists

Check how each framework is described, what it can and cannot measure, and where its evidence ends.

Psychiatry

Psychiatrists

Make sure no finding reads as a diagnosis, and that anything touching on risk points to real care.

Counseling

Counselors & therapists

Read reports the way a client would, and flag wording that could mislead, shame or escalate.

What they review

  1. 01

    Framework review

    How each framework is summarized, including its known weaknesses.

  2. 02

    Language review

    How findings are worded, so a pattern never reads as a verdict on a person.

  3. 03

    Safety review

    What a report says when a conversation suggests someone may be at risk, and where it points them.

  4. 04

    Scenario testing

    Reports generated from synthetic, made-up conversations are read and critiqued before changes ship.

Our clinical advisors are independent licensed professionals who are paid for their time. Their review shapes how MOSAIC is built and worded. It is not an endorsement of any individual report, and advisors never see your conversations or reports.

Foundations

The research we build on.

Publications
6,661
Years of research
154
Fields
15
  1. 2026

    “I can feel that you love me”: Approximating affectionate touch in long-distance romantic relationships

    Mitala, J., Dickerson, S., & Jakubiak, B. K. · Journal of Social and Personal Relationships

    Cited in MOSAIC’s care languages guide.

    CareView
  2. 2026

    A historical review of gaslighting: Tracing changing conceptualizations within psychiatry and psychology

    Klein, W., Wood, S., Forget, A.-A., & Bartz, J. A. · Clinical Psychology Review, 126, 102742

    Cited in MOSAIC’s manipulation tactics guide.

    NarcissismView
  3. 2026

    A Theoretical Framework for Studying the Phenomenon of Gaslighting

    Klein, W., Wood, S., & Bartz, J. A. · Personality and Social Psychology Review, 30(2), 195–215

    Cited in MOSAIC’s manipulation tactics guide.

    NarcissismView
  4. 2026

    Adaptive humor styles as predictors of post-traumatic growth factors

    Kruger, G. · Behavioral Sciences, 16(7), 1131

    Cited in MOSAIC’s humour styles guide.

    TraitsView
  5. 2026

    An ecological text message experiment of Hispanic and Latino adolescents' recognition of and emotional responses to digital dating abuse behaviors

    Figueroa, J. M., Ha, T., Anderson, S. F., Maras, O., & Quiroz, S. I. · Journal of Research on Adolescence, 36(1), e70167

    Cited in MOSAIC’s manipulation tactics guide.

    NarcissismView
  6. 2026

    Are psychologists aware of group psychological abuse? A survey on awareness of the phenomenon among Italian psychologists

    Tinelli, L., Destratis, P., Curci, A., Schincariol, A., Melis, G., & Scarpazza, C. · Frontiers in Psychology, 17, 1749074

    Cited in MOSAIC’s manipulation tactics guide.

    NarcissismView
  7. 2026

    Assessing Perpetrator Responses to Confrontation: Associations with a DARVO-SF and Posttrauma Symptoms in Two Different Populations

    Durland, M., Harsey, S. J., & Freyd, J. J. · Journal of Trauma & Dissociation, 27(2), 221–237

    Cited in MOSAIC’s manipulation tactics guide.

    NarcissismView
  8. 2026

    Assessing the personality traits of mid- and senior-level nurses

    Koç Aslan, S. · Croatian Nursing Journal, 10(1), 51–62

    Cited in MOSAIC’s DISC guide.

    Enneagram · DISCView
  9. 2026

    Attachment styles and defense mechanisms as predictors of marital satisfaction in Iranian couples

    Nejad, F. K., Mahmoudi, M., Hiebler, M., & Unterrainer, H. F. · Frontiers in Psychology, 17, 1659840

    Cited in MOSAIC’s defense styles guide.

    EmotionView
  10. 2026

    Automated scoring of adult ego development from sentence completions: Interpretable maturity indicators derived by evolutionary prompt optimization

    Madl, T., & Cook-Greuter, S. R. · Frontiers in Psychology, 17, 1853871

    Cited in MOSAIC’s ego development guide.

    DevelopmentView

Listing a publication means MOSAIC’s approach draws on its ideas. It does not mean its authors are affiliated with, reviewed or endorse MOSAIC.

Honest limits

What the research won’t let us say.

Accuracy score

Not published

You won’t find an accuracy percentage on this site. We won’t publish one until it comes from a completed, independent validation study.

  1. 01

    Text is only part of a person

    Messages miss tone of voice, context and everything said face to face. A report reflects the conversations you chose, not the whole relationship.

    How conversations are chosen
  2. 02

    A pattern is not a diagnosis

    Only a qualified clinician who assesses someone directly can diagnose a condition such as a personality disorder. MOSAIC never does, and never labels a person.

    Read the FAQ
  3. 03

    AI can be wrong

    Language models can misread sarcasm, in-jokes, culture and second languages. That is why every finding shows its evidence and its confidence.

    See how findings show evidence
  4. 04

    Frameworks have limits too

    Some popular models, such as MBTI, have well-documented weaknesses. We describe those limits inside the report instead of hiding them.

    The personality lens

See it for yourself

Research you can read yourself.

Every finding in your report shows the messages behind it and how confident MOSAIC is. Read the evidence and decide for yourself.