AI Emotion Detection —
Three Sources, One Truth
Emotions you feel but cannot name are often the most important ones. PsyChi finds them anyway.
Human self-reporting of emotion is unreliable — not because people lie, but because emotional awareness is partial by nature. The most clinically significant feelings are frequently the ones least consciously accessible. PsyChi addresses this with a three-source emotion detection model that cross-validates what you say against how you say it and what your voice reveals physiologically.
Source 1 — Verbal ContentSource 1 — Verbal Content
What you explicitly say. If you name an emotion directly, the AI captures it and rates its intensity on a 0–5 scale based on the language and context of your entry. Verbal content carries a source weight of 1.0 — the highest of the three sources.What you explicitly say. If you name an emotion directly, the AI captures it and rates its intensity on a 0–5 scale based on the language and context of your entry. Verbal content carries a source weight of 1.0 — the highest of the three sources.
Source 2 — Qwen Audio Tone AnalysisSource 2 — Qwen Audio Tone Analysis
How you say it. Qwen's audio AI analyzes the prosodic qualities of your voice in real time — pace, pitch, tension, and hesitation. These vocal patterns carry emotional information that exists independently of your word choice. Qwen converts its confidence reading into an intensity score on the same 0–5 scale, with a source weight of 0.75.How you say it. Qwen's audio AI analyzes the prosodic qualities of your voice in real time — pace, pitch, tension, and hesitation. These vocal patterns carry emotional information that exists independently of your word choice. Qwen converts its confidence reading into an intensity score on the same 0–5 scale, with a source weight of 0.75.
Source 3 — EVI & ASR Emotion AISource 3 — Hume
Emotion AI
What your voice reveals beneath the surface. Hume's physiological model reads micro-patterns in vocal production — breath rhythm, resonance, subtle muscle tension — that correlate with specific emotional states at a level below conscious awareness. This is the deepest layer of detection, also weighted at 0.75.
The three sources produce a single combined intensity score per emotion. Because they are independent, they cross-validate: when all three agree, the AI has high confidence. When they diverge, the system captures the discrepancy — which is itself psychologically meaningful.What your voice reveals beneath the surface. Hume's physiological model reads micro-patterns in vocal production — breath rhythm, resonance, subtle muscle tension — that correlate with specific emotional states at a level below conscious awareness. This is the deepest layer of detection, also weighted at 0.75.
The three sources produce a single combined intensity score per emotion. Because they are independent, they cross-validate: when all three agree, the AI has high confidence. When they diverge, the system captures the discrepancy — which is itself psychologically meaningful.
Monthly Miracles Reports — Progress Made VisibleMonthly Miracles Reports — Progress Made Visible
Once a month, everything you have healed — shown clearly.
The Miracles Report is where PsyChi's methodology produces its most tangible output. Each month, the AI runs a complete analysis of your journal history and generates a structured report showing how your inner world has shifted.Once a month, everything you have healed — shown clearly. The Miracles Report is where PsyChi's methodology produces its most tangible output. Each month, the AI runs a complete analysis of your journal history and generates a structured report showing how your inner world has shifted.
Thread ClusteringThread Clustering
Once a month, everything you have healed — shown clearly.
The Miracles Report is where PsyChi's methodology produces its most tangible output. Each month, the AI runs a complete analysis of your journal history and generates a structured report showing how your inner world has shifted.
Related LoMs under each Path are grouped into miracle threads — thematic clusters of fear or love patterns sharing a common psychological root. Clustering uses cosine similarity: LoMs scoring 0.70 or above on the similarity measure are assigned to the same thread. Each thread is named after its founding fear pattern.Once a month, everything you have healed — shown clearly.
The Miracles Report is where PsyChi's methodology produces its most tangible output. Each month, the AI runs a complete analysis of your journal history and generates a structured report showing how your inner world has shifted.
Related LoMs under each Path are grouped into miracle threads — thematic clusters of fear or love patterns sharing a common psychological root. Clustering uses cosine similarity: LoMs scoring 0.70 or above on the similarity measure are assigned to the same thread. Each thread is named after its founding fear pattern.
Ascent and Descent ClassificationAscent and Descent Classification
Each thread is classified monthly as Ascent (love accumulating faster than fear) or Descent (fear still dominant). Classification uses a weighted scoring formula accounting for recency, frequency, and detection source of each LoM.Each thread is classified monthly as Ascent (love accumulating faster than fear) or Descent (fear still dominant). Classification uses a weighted scoring formula accounting for recency, frequency, and detection source of each LoM.
Hysteresis — Protecting Real ProgressHysteresis — Protecting Real Progress
A difficult week does not erase months of healing. PsyChi applies hysteresis rules requiring sustained evidence across multiple entries before a thread classification changes. The report reflects durable transformation — not mood fluctuation.A difficult week does not erase months of healing. PsyChi applies hysteresis rules requiring sustained evidence across multiple entries before a thread classification changes. The report reflects durable transformation — not mood fluctuation.
The Switched BadgeThe Switched Badge
When a thread completes a confirmed shift from Descent to Ascent, it receives a Switched badge — visible for two months. This is a miracle in the ACIM sense: a shift in perception from fear to love, now documented and dated.When a thread completes a confirmed shift from Descent to Ascent, it receives a Switched badge — visible for two months. This is a miracle in the ACIM sense: a shift in perception from fear to love, now documented and dated.
Three Lesson Applications Per ThreadThree Lesson Applications Per Thread
For every active thread, the AI generates three personalized ACIM lesson applications — specific instructions for working with that exact fear pattern using the current month's lessons. Not affirmations. Not generic advice. Tailored to your history, your path, and where you are in your three-year journey.For every active thread, the AI generates three personalized ACIM lesson applications — specific instructions for working with that exact fear pattern using the current month's lessons. Not affirmations. Not generic advice. Tailored to your history, your path, and where you are in your three-year journey.