Our Personality Insights and the Applications of AI

Dr. Martin Lukac

Co-Founder & Chief Data Scientist

Published on —

October 11, 2023

4 min read



Introduce Koios and what we offer

Unlike traditional personality assessments, our Koios system offers a time-efficient and simple method that provides nearly instantaneous results for our users. Not only that, but our personality assessment provides each user with personalised insights that are tailored based on the acoustic and linguistic features analysed by our AI model from the voice data we receive. 

About our personality insights

Our Koios personality assessment method is based on a carefully validated psychometric instrument for measuring personality traits along with data sorting methods, which when tailored to measure acoustic data revealed that people tend to fall into 25 distinct personality clusters. 

So, when users engage with our platform, they will discover which of the 25 personality types, revealed through our in-depth research, most accurately characterises them.

To learn more about how we specifically trained our AI model and generated our 25 personality types, see here.

What is special about our personalised insights?

At its core, our service is a combination of a data-driven calculation of someone’s Big Five personality score based on the voice biomarkers extracted from the acoustic signals within each user’s voice recording AND the linguistic content of what each user actually discussed in their voice recording. 

This way, our users will not simply receive a generalised personality assessment description of the personality type they fall under  – every report is personalised based on the contents of their voice recording, allowing for an even more accurate and truly distinctive representation of each of our users. 

Let’s take a closer look at how our AI system functions by examining both the linguistic and acoustic aspects of our analysis:

What is being said: Linguistics

The initial step in our model's processing involves cleaning and preprocessing linguistic signals from audio recordings by removing background noise and irrelevant sounds. These cleaned signals are then transcribed using automatic speech recognition (ASR) technology. Subsequently, linguistic and contextual features are extracted, such as word choice, sentence complexity, and context inferred from speech, enabling us to correlate linguistic patterns with personality traits. This information is used to create a linguistic dataset, a crucial input for our personality prediction models.

How it is being said: Acoustics

Our acoustic signal processing involves normalising raw audio, cleaning and preprocessing to remove unwanted sounds, and extracting voice biomarkers scientifically linked to personality traits. Additionally, pre-trained audio models extract various acoustic features, creating an acoustic dataset that complements the linguistic dataset. These combined datasets provide a comprehensive understanding of the speaker, contributing to our final personality score predictions.

To learn more about our AI model, see here.

Ultimately, what makes our models so special is that they are able to capture the nuance and depth of human personality that cannot be achieved with a traditional self-assessment or questionnaire-based method. Each personality assessment is a testament to the unique blend of what is being said and how it is being said, offering an innovative and truly comprehensive depth of insight and accuracy into the scope of human personality.