Our Analysis Core
Deception detection has existed for decades, but it has only now began to be empowered by AI and neural networks. We aim to be at the forefront of these developments, which are set to disrupt past beliefs/approaches.
-
Computer Vision
Extracts facial behaviors and nonverbal cues from video recordings using advanced visual algorithms.
-
Audio Spectrograms
Converts audio into mel spectrograms to analyze pitch, tone, and speech dynamics.
-
Transcription
Automatically transcribes speech to text for detailed linguistic analysis and statement-by-statement segmentation.
-
LLM Analysis Interpretation
Uses large language models to interpret data and generate human-friendly analytical summaries.
Our Approach
-
1. PERCEIVE
We monitor visual, audio, and linguistic patterns to capture behavioral cues across multiple communication channels.
-
2. LEARN
Algorithms automatically establish a consistent initial behavioral baseline pattern for a person during the first 10–30 seconds of recording.
-
3. DETECT
We promptly raise an automated alert whenever we detect unusual behavioral patterns or covert cues linked to deception attempts.
-
4. PROCESS
We then segment our analysis into individual conversational statements, allowing rapid, targeted evaluation of each statement’s significance.
-
5. EVALUATE
We calculate a risk score for each segmented conversational statement using all collected behavioral data and evaluation criteria to generate actionable insights.
Cues we're looking For
Visual Modality
Heart Rate
Rapid Spikes & Drops
Blinking
Altered blink rate frequency
Eye Gaze
Shifted eye gaze direction
Risk Emotions
Detected elevated risk emotion
Hands on Face
Hand touches face
Head Orientation
Altered head orientation angle
Looked Up/Down
Glance up or down
Audio Modality
Pitch Changes
Sudden pitch spike
Volume Changes
Rapid Spikes, Drops
Speech Rate Changes
Rapid speech variation
Extended Pauses
Lengthy silent gap
Linguistic Modality
Risk Words
Risk-indicating words
Filler Words
Frequent filler usage
Suspicious Phrases
Abnormal phrase patterns
Negation Use
Excessive negative words
Psychological Modality
MBTI Personality Type
Personality classification system
Attachment Style
Patterns of relational bonding
Light/Dark Triad
Spectrum of personality traits
Deception Detection in Academia
Start with Kley by signing up and connecting your accounts. Enjoy personalized budgets, real-time tracking, and insightful analytics to take control of your finances and achieve your goals.
- See for yourself
See our Technology in Action
With our web application you can upload and analyze any video you want, or you can send out AI-powered interviews
Social Impact and Privacy Stance
We do not collect user data for training purposes and we are striving to make a positive impact with our technology
FAQs
What is PolygrAI and how does it work?
How accurate is PolygrAI?
What behavioral and physiological cues does PolygrAI analyze?
How does PolygrAI detect deepfakes?
What psych profile insights does PolygrAI provide?
How is my privacy protected?
How do you calculate the AI Score?
Web and Mobile Access?
What integrations are supported?
DISCLAIMERS
PolygrAI will never reach absolute accuracy and it's illegal to use without consnet
While our technology provides valuable insights into potential risks, it’s crucial to understand that human behavior is inherently complex and unpredictable. Our system is designed to assist users by offering data-driven assessments that aid interpretation, but it should not be viewed as a conclusive determination of truth.
Users are encouraged to consider PolygrAI’s analysis as one part of a broader evaluation process, complementing other information and judgment. Always prioritize ethical use and respect individual privacy and rights.
Yet, it can still make a difference
Despite inherent limitations, our ongoing research and development efforts are focused on refining our system’s accuracy and utility to better serve our users and enhance its impact.
As an AI-powered, multimodal platform under active development that integrates more behavioral and physiological cues than traditional polygraphs, we expect our technology to surpass conventional systems over time, offering increasingly reliable insights when used alongside human judgment.
