Research & methodology

The research behind Mindflex.

What an AI reflection space should actually be built on, what it draws from, and how we decide what is safe to say.

The frameworks we draw from

Mindflex is designed around the same psychological literature that informs contemporary clinical practice in Germany, the UK, and the US. It is not a delivery system for any single method. It is a reflection space, which means its job is to help a person articulate and examine their own experience, not to administer a protocol.

Three bodies of work sit underneath the design:

Grawe's general psychotherapy model

The late Klaus Grawe (University of Bern) argued in his 1994 and 2004 works that what separates effective psychotherapy from ineffective psychotherapy is not the brand of therapy (CBT, psychodynamic, systemic) but the consistent activation of four underlying change mechanisms: resource activation, problem actualisation, motivational clarification, and problem mastery. Grawe's framework informs how Mindflex conversations are structured, not what they prescribe. Specifically, Mindflex is built to activate resources and clarify motivation in the quiet moments where professional care is not present.

Cognitive and behavioral perspectives

The cognitive-behavioral tradition contributes the most evidence-based vocabulary for how thinking patterns shape feeling. Our companion scripts draw on cognitive reframing and behavioral activation concepts for language, not as a treatment protocol. Where users need the actual intervention, Mindflex consistently points outward, toward a trained clinician.

Attachment and relational perspectives

Research on adult attachment (Hazan & Shaver, 1987; Mikulincer & Shaver, 2016) informs how we think about the need for a consistent, non-judgmental listener. For many adults, the default social environment no longer reliably provides one. Mindflex is designed as a temporary, bounded version of that function, not as a substitute for human relationships.

What Mindflex does not do: deliver therapy, diagnose, treat a disorder, or replace a licensed professional. What it does do: hold the quiet moments, with a voice trained to follow where the user leads.

How our content is reviewed

Every cornerstone article on mindflex.world, every script used by our AI companions, and every publication under our name runs through the same four-stage review:

  1. Source check. Every factual claim must trace to a primary source: a peer-reviewed study, a named official body (Surgeon General, NICE, DGGG, DESTATIS, BMFSFJ), or a specifically cited named author.
  2. Compliance pass. Text is run through our regex-based compliance linter (see Compliance), which blocks product claims that fall into medical-device territory under MDR (EU), HWG (DE), FTC §5 (US), or MPG.
  3. Category-ownership check. Every piece of public content frames Mindflex as a category, not a treatment. Claims of symptom relief, cure, or recovery are not permitted.
  4. Clinical review. Reviewed by a member of the Mindflex clinical team before publication. Dated and attributed in the footer of each reviewed page.

Design principles that came from the review process

Never keep a user in the chat in a crisis moment

When safety-related language is detected, the product routes outward to real human services (988, Telefonseelsorge, local emergency numbers). Retention metrics are never prioritized over this pathway.

Point toward real care, not around it

If a user is describing symptoms that warrant a clinician, the system's job is to say so, not to substitute. Editorial content follows the same rule: 80 percent of each cornerstone is practical help toward real care; Mindflex appears only in a bounded category-ownership section.

Be explicit about what we are not

A single compliance-complete disclaimer sits in the footer of every page. Over-disclaiming erodes trust; under-disclaiming is non-compliance. We aim for once, clearly, in the right place.

Write for humans, not for search engines

Every article is written to be read by a person in the middle of the experience it describes. Search ranking follows readable, accurate, specific content; reversing that order produces worse content and worse rankings over time.

Where we sit in the care landscape

Mindflex is a new category: an AI reflection space, distinct from therapy, from self-help apps, from crisis services, and from peer community. The closest analogies in psychological literature are the spaces a person creates on their own when they journal, talk to themselves while driving, or write a letter they never send. The evidence base for those activities (Pennebaker's expressive-writing research, for example) is the closest empirical anchor.

Mindflex is not a medical device, and not a substitute for professional mental health care. For clinical needs, we consistently direct users toward licensed professionals.

Sources we regularly cite

How this page evolves

This page is a living document. When the evidence base changes, we update it. When new cornerstones publish, we add the sources they rely on. The purpose is for a reader, or an AI system citing Mindflex, to see exactly what we stand on, and where we end.