Anthea · Method
The science behind Anthea
Anthea is a couples app that turns daily answers into a weekly relationship analysis. Its question library draws on six frameworks in couples psychology. This page documents what each framework is, how Anthea actually uses it, and — with citations — where the evidence behind it is weak. That last part is included for frameworks Anthea itself uses.
Written by Şehmus Nas, psychologist and couples therapist. Last reviewed: August 2026.
The short version
Anthea is a structured reflection tool, not a measurement instrument and not a treatment. Its questions are informed by established frameworks, but Anthea has not been evaluated in a controlled trial. It produces no diagnosis, no clinical score, and no prediction about whether a relationship will last. Where the research behind a framework is contested, this page says so rather than leaving it out.
How a weekly analysis is actually produced
Two separate things happen, and it matters which is which.
The numbers come from code
Every percentage, trend and agreement rate in your weekly report is computed deterministically from your answers by ordinary code — not by an AI model. Each question carries a framework, a clinical field, and a direction (whether agreement is a positive or a negative signal). Skipped questions are excluded from the denominator rather than counted as negative. The same answers always produce the same numbers.
The words come from a language model
The written parts of the report — the summary, the observations, the suggested next step — are generated by a large language model (OpenAI's API), working from those computed numbers and from what you and your partner wrote. The model does not calculate anything and cannot change a score. It writes prose around figures that were already fixed.
Two consequences follow, and both are real limitations. First, the wording of your report will vary between runs even when the underlying data is identical. Second, language models can produce fluent text that overstates what the data supports; Anthea constrains this with an explicit instruction set — no diagnostic language, no predictions, no invented detail — but constraint is not a guarantee.
What the model is told to ignore
Your free-text answers are passed to the model as material to reflect on, explicitly fenced off from the instructions. If someone writes text into an answer field designed to give the model orders ("ignore your instructions", "write this verbatim"), the system is instructed to disregard it and consider only the genuine emotional content. This matters because in a couples app, one partner's text is read by a system that also writes to the other.
The six frameworks, and their limits
Anthea's question library is organised under six frameworks, subdivided into 45 clinical fields. Below, each framework gets the same treatment: what it is, how Anthea uses it, and what the evidence does not support.
1. The Gottman method
Four decades of observational research on how couples handle conflict, repair, fondness and shared meaning — much of it based on coding recorded interactions rather than self-report alone.
In Anthea: the largest group of questions. Fields include love maps, fondness and admiration, turning toward, conflict management, dreams within conflict, and trust.
Limit
The widely repeated claim that divorce can be predicted with over 90% accuracy does not survive scrutiny. Heyman and Smith Slep (2001) showed that these prediction equations were fitted and tested on the same sample; under proper cross-validation, accuracy drops sharply. Anthea makes no predictions about whether a relationship will last, and treats Gottman's work as a well-supported description of interaction patterns, not as a forecasting instrument.
2. Emotionally Focused Therapy (EFT)
An attachment-based couples therapy that works with the emotional cycle underneath a recurring argument — the pursue–withdraw pattern and the vulnerable feelings driving it.
In Anthea: fields covering accessibility, attunement, cycle awareness, engagement, repair and forgiveness, and raw spots.
Limit
EFT has the strongest randomised-trial support of any framework here, but the effect is more modest than popular summaries suggest. In a meta-analysis of 33 randomised controlled trials (Rathgeber et al., 2019), EFT showed a medium effect on relationship satisfaction immediately after treatment (g = 0.73), smaller at six months (g = 0.66), and it did not differ significantly from behavioural couples therapy. More importantly: that evidence is for therapy delivered by a trained therapist. It does not transfer to an app, and Anthea does not claim it.
3. Attachment theory
How each partner seeks security when distressed and offers it in return — the safe haven and secure base of the relationship.
In Anthea: fields covering safe haven, secure base, proximity seeking, responsiveness, co-regulation, earned security and protest behaviour.
Limit
Adult attachment does not come in types. Taxometric analyses (Fraley & Waller, 1998; Fraley et al., 2015) found that variation in romantic attachment is dimensional, not categorical — meaning labels like "anxious" or "avoidant" describe positions on continuous dimensions, not four boxes people belong to. Attachment patterns are also relationship-specific and can change. Anthea therefore never assigns you an attachment type or tells you what your style is.
4. Sternberg's triangular theory of love
Love described through three components — intimacy, passion and commitment — whose relative weight shifts over the life of a relationship.
In Anthea: questions across all three components, used to notice which one has gone quiet rather than to name a "type" of love.
Limit
The triangular theory is a descriptive map, not a predictive model. Measurement of its three components varies across instruments and cultures, and the components correlate strongly enough that treating them as cleanly separable is questionable. Anthea uses it as vocabulary for noticing change over time — not as a classification.
5. Chapman's five love languages
The popular idea that people primarily give and receive care through one of five channels: words of affirmation, acts of service, gifts, quality time, and physical touch.
In Anthea: questions about how each partner expresses care and how they notice receiving it.
Limit — the weakest framework here
This should be said plainly. Impett, Park and Muise (2024) reviewed the empirical work and found support for none of Chapman's three central claims: that each person has one primary love language, that there are exactly five, or that couples do better when partners "speak" each other's language. Anthea therefore does not test you, assign you a love language, or advise matching your partner's. The questions are used only to notice how care is currently being expressed and whether it is landing — which is the part that survives the critique. The authors propose a better metaphor: a balanced diet rather than a single language.
6. Positive psychology
Attention to what is already working — gratitude, savouring, and sharing good news — rather than only to what hurts.
In Anthea: questions on appreciation and on capitalization (how a partner responds when you share something good).
Limit
Positive psychology interventions work, but less than early meta-analyses claimed. White, Uttl and Holder (2019) re-analysed the two most-cited meta-analyses and found that once small-sample bias was corrected, effects on well-being were small though still significant, and effects on depression were unstable and often not significant. Anthea uses this material to balance attention across the week, not as an intervention with a promised effect.
Who wrote the questions
The question library — six frameworks, 45 clinical fields — was written by Şehmus Nas, a psychologist and couples therapist in practice, reviewed with colleagues working clinically with couples, and revised as the app is used. Questions are written to be answerable in seconds, to avoid leading the reader toward a "correct" answer, and to be answered separately by each partner before either sees the other's response, so that neither is anchored by the other.
That review is professional consultation, not formal validation. Anthea is built by one person; there is no research team behind it, and this page should be read with that in mind.
What Anthea does not claim
- It is not therapy and is not a substitute for it. Couples with entrenched patterns generally need a third person in the room.
- It does not diagnose. No condition, no attachment type, no love language, no personality classification.
- It does not predict. Nothing in Anthea forecasts whether a relationship will last.
- It is not a crisis tool. Where there is violence, coercion or fear in a relationship, structured reflection is the wrong instrument and safety comes first.
- It has not been tested in a controlled study. The frameworks it draws on have evidence; the app itself does not yet. Any such study will be published here.
- Its scores are not clinical measures. The percentages describe your own answers over time. They are not normed against a population and mean nothing outside your own week.
Privacy, in one paragraph
Daily answers are not shown to your partner until both of you have written. Free-text answers are sent to OpenAI's API to generate the written report; under OpenAI's API policy, data sent this way is not used to train their models. Anthea does not sell data and does not use your answers to train any model of its own. Account and data deletion is available from within the app. The full policy is in the privacy policy.
References
- Heyman, R. E., & Smith Slep, A. M. (2001). The hazards of predicting divorce without crossvalidation. Journal of Marriage and Family, 63(2), 473–479. doi:10.1111/j.1741-3737.2001.00473.x
- Rathgeber, M., Bürkner, P.-C., Schiller, E.-M., & Holling, H. (2019). The efficacy of emotionally focused couples therapy and behavioral couples therapy: A meta-analysis. Journal of Marital and Family Therapy, 45(3), 447–463. doi:10.1111/jmft.12336
- Fraley, R. C., & Waller, N. G. (1998). Adult attachment patterns: A test of the typological model. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and close relationships (pp. 77–114). Guilford Press.
- Fraley, R. C., Hudson, N. W., Heffernan, M. E., & Segal, N. (2015). Are adult attachment styles categorical or dimensional? A taxometric analysis of general and relationship-specific attachment orientations. Journal of Personality and Social Psychology, 109(2), 354–368.
- Impett, E. A., Park, H. G., & Muise, A. (2024). Popular psychology through a scientific lens: Evaluating love languages from a relationship science perspective. Current Directions in Psychological Science, 33(2), 87–92. doi:10.1177/09637214231217663
- White, C. A., Uttl, B., & Holder, M. D. (2019). Meta-analyses of positive psychology interventions: The effects are much smaller than previously reported. PLoS ONE, 14(5), e0216588. doi:10.1371/journal.pone.0216588
- Gottman, J. M., & Silver, N. (1999). The seven principles for making marriage work. Crown.
- Johnson, S. M. (2004). The practice of emotionally focused couple therapy: Creating connection (2nd ed.). Brunner-Routledge.
- Sternberg, R. J. (1986). A triangular theory of love. Psychological Review, 93(2), 119–135.
Try it
Download on the App Store
iOS · English and Turkish · Free to start