The Science Behind the Platform
1.
Measure Development and Data Collection
Selecting or creating the right data to train AI algorithms is essential and should be one of the primary considerations when developing AI. Training an AI algorithm to learn how to rate the psychological characteristics of a speaker requires high-quality and real-world speech data along with accurate ratings by expert human raters. The first step in producing such data was to collect high-quality recordings of natural real-world speech and to develop valid psychological measures of personality traits, mindset, and emotional states.
2.
Rater Selection and Training
The second step in producing high-quality speech data was to hire and train a group of experts in psychological assessment who would provide the ratings for the AI to learn from. A large team of individuals with educational training and professional experience in psychological assessment was selected to provide these ratings. These individuals received training on best practices for speech-based psychological assessment.
.png)
3.
Rater Development and Agreement
The third and final step in creating high-quality data for the AI to learn from involved rating each speaker in each audio file using the new measures of personality traits, mindset, and emotional states. The group of raters completed these ratings using a propriety rating application developed by VoiceSignals. All raters received ongoing monitoring and feedback. Statistical analyses showed that the ratings provided by the raters had very high levels of agreement and reliability.
AI Development
Segments of speech with high levels of rater agreement were selected for training the AI. The AI development process was thorough and aligned with industry best practices, which included using transparent algorithms, multiple stages of validation testing, and the investigation and mitigation of bias. The final AI algorithms displayed excellent levels of accuracy in predicting a speaker's personality traits, mindset, and emotional states.
AI Agreement
After the AI had been developed, a series of validation studies were conducted following the standards for psychological assessment. Results showed that the scores produced by the AI algorithms embedded in the People Intelligence Platform© are highly valid predictors of a speaker's personality traits, mindset, and emotional states.
Learn more as our platform learns with you.
Match
Match voices with key personas/behavioral profiles that can provide ever more accurate predictions of future behavior.
Monitor
Monitor important changes in an individual or group’s behavior so you can communicate the right way at the right time.
%201.svg)
Moderate
Moderate changing emotional states over the course of a conversation to boost your emotional intelligence and build stronger relationships.
Latest posts
%201.svg)
