Dialectic Method is built around generative AI. We believe in being explicit about what the AI does, what it does not do, and what the limits are.
What the AI does
- Generates the interlocutor's spoken replies across the six dialectical stages (Proposition → Reflection).
- Scaffolds within your Zone of Proximal Development — adjusting difficulty using CEFR / ACTFL targets.
- Produces rubric-style feedback across range, accuracy, fluency, interaction, and coherence.
- Estimates your CEFR level and ACTFL task success based on transcript evidence. These are estimates, not certifications.
What the AI does not do
- It does not issue official language certifications.
- It does not provide legal, medical, financial, or psychological advice — even when topics touch those domains.
- It is not a human teacher and does not replace one.
Models and providers
We route AI traffic through enterprise gateways to leading model providers (e.g. Google Gemini, OpenAI GPT). Provider contracts prohibit training of general-purpose models on your content. Specific model selection may evolve as quality and safety improve.
Voice and speech
Speech recognition uses your browser's on-device speech engine where available. Synthesized voice output uses your device's text-to-speech engine; premium server-side voices may be introduced in future and will be disclosed before activation.
Limitations and bias
Large language models can produce inaccurate, inconsistent, or culturally biased framings — particularly on contested philosophical, political, and ethical topics. Treat AI replies as a sparring partner for argumentation, not as authoritative truth.
Human oversight
Pedagogical design, rubrics, scaffolding logic, and topic curation are designed and maintained by humans. We monitor aggregate quality signals and welcome feedback at feedback@dialecticmethod.com.
Your control
- You may delete any session from your history.
- You may export all your data from Settings → Privacy.
- You may permanently delete your account at any time.