Thesis
Behavior is observable. Intent is not.
WizeLens is built on a narrow, defensible claim: with enough video and the right frameworks, a trained observer can describe a person's behavior — their posture, expressions, voice, and gaze — far more consistently than the unaided eye. Those descriptions are useful to coaches and recruiters because they replace gut feel with a structured record.
What we don't claim is anything beyond the surface. We don't infer truthfulness, character, employability, or mental state. We don't produce a "lie score." We don't read minds. Every number we surface describes something the camera and microphone actually captured.
Our north star: every observation must be traceable to a frame, a waveform, or a timestamp. If we can't point at it in the recording, we don't score it.
The four channels
What we measure, and how
Every WizeLens analysis runs four independent signal channels in parallel. Each channel produces its own observations and sub-scores before anything is combined.
Body language
Pose estimation tracks shoulders, head, hands, and torso across the recording. We measure posture stability, gesture frequency and openness, mirroring with other speakers, and self-soothing behaviors (face touches, neck rubs).
Research basis: Builds on the nonverbal taxonomy work of Joe Navarro and the broader proxemics/kinesics literature.
Micro-expressions
Frame-level facial analysis detects action units associated with the basic affect families (joy, sadness, anger, fear, disgust, surprise, contempt) and reports their intensity, frequency, and timing.
Research basis: Grounded in Paul Ekman's Facial Action Coding System (FACS) and subsequent affect-detection research.
Vocal tone & prosody
Audio analysis extracts pitch contour, pace (words per minute), pause structure, intensity dynamics, and vocal fry. We summarize range, variability, and where conviction rises or drops.
Research basis: Draws on affective-computing and prosody research; conviction cues per Bowden's communication frameworks.
Eye contact & gaze
Gaze estimation tracks where the subject is looking relative to the camera, plus blink rate and aversion patterns. We report sustained-contact ratios and the timing of breaks.
Research basis: Standard gaze-tracking literature; interpretation is descriptive only.
Scoring framework
From observations to the Behavioral Signal Score™
Each channel produces normalized sub-scores on a 0–100 scale. The headline Behavioral Signal Score is a weighted composite — intentionally boring, intentionally transparent. The weights below are the current defaults; coaches and recruiters can override them per template.
Default weights
- Composure20%
- Congruence20%
- Vocal conviction15%
- Eye-contact consistency15%
- Affect range15%
- Gesture openness15%
How a sub-score is built
- Raw signal extraction — frame-level features (pose, AUs, pitch, gaze) are computed for the full recording.
- Temporal aggregation — features are summarized in rolling windows so brief outliers don't dominate.
- Normalization — values are scaled against a reference distribution drawn from comparable recordings (interview, coaching, presentation).
- Evidence linking — every sub-score is paired with the timestamps that drove it, so reviewers can verify the call.
A score without evidence is just an opinion. Every number in a WizeLens readout links back to the moments that produced it.
Honest limits
What WizeLens can — and cannot — infer
Behavioral video analysis is genuinely useful and routinely overclaimed. Here is the line we draw, in plain language.
What it can do
- Describe posture, gesture, gaze, and vocal patterns with consistent, timestamped evidence.
- Surface moments of high or low expressive variability worth reviewing.
- Compare a person's delivery across multiple sessions on the same metric set.
- Give coaches and recruiters a shared, structured vocabulary for feedback.
What it will not do
- Detect deception, deceitfulness, or whether someone is 'lying'.
- Diagnose mental health, neurodivergence, or medical conditions.
- Predict job performance, hireability, or future success.
- Infer character traits, values, or 'who someone really is'.
- Account for cultural, neurodivergent, or disability-related differences in nonverbal expression — interpretation always belongs to the human reviewer.
Nonverbal behavior varies meaningfully across culture, neurotype, and context. Scores are a starting point for human judgment, never a substitute for it.
Ethics & data
The rules we hold ourselves to
- Consent first. Every analysis requires explicit confirmation that the person on camera has agreed to be analyzed.
- Description over diagnosis. Scores describe observable behavior. They never assign character, intent, or clinical labels.
- Evidence on demand. Every number is traceable to a timestamp the reviewer can replay and override.
- No lie detection. We do not, and will not, ship deception-detection features. The science doesn't support them and the harm potential is high.
- Privacy by default. Recordings are encrypted in transit and at rest, scoped to the account that uploaded them, and deletable on request.
Coaches and recruiters carry the final judgment. WizeLens carries the observations.
Research lineage
The shoulders we stand on
WizeLens does not invent a new theory of human behavior. It operationalizes well-established frameworks and is honest about where they end.
Paul Ekman
Emotions Revealed; Facial Action Coding System (FACS)
Foundational taxonomy for facial affect and micro-expressions.
Joe Navarro
What Every BODY Is Saying
Nonverbal behavior frameworks from observation in high-stakes interviews.
Vanessa Van Edwards
Cues
Practical synthesis of nonverbal, vocal, and verbal signals.
Mark Bowden
Winning Body Language
Gesture-plane and conviction frameworks for delivery.
Daniel Goleman
Emotional Intelligence; Social Intelligence
Affect-recognition and interpersonal-attunement foundations.
Albert Mehrabian
Silent Messages
Often-misquoted source for the nonverbal share of emotional communication.
See the framework in action.
Upload a video and watch every score link back to the moments that produced it.
Try a free analysis