Is it possible to use AI to measure people's mental state?

In 2019, the two co-founders of Erudit AI, Alejandro Martínez and Ricardo Michel, asked themselves this question while looking at the IBM Watson feature catalog. Its Personality Insights API provide a Big 5 personality analysis model based on:

  • Openness to new experiences (curious / cautious)
  • Discipline (organized / careless)
  • Extroversion (energetic / reserved)
  • Cooperativity (friendly / detached)
  • Neuroticism (nervous / confident)

Intrigued by the model, they created a minimum viable product (MVP) based on it, which they later called Hero Talent, in order to see whether the market was interested in knowing the personality of employees. Using this product, employees could log into the platform with their email addresses and the tool could compile a dashboard with the 5 personality dimensions based on Watson’s model.

Fortunately, the MVP had a positive response and we found that other companies, such as Crystal Knows, were already offering similar services in the market. While most new market entrants were weary of the existing competition, the co-founders of Erudit were relieved to find alternatives in the market. It meant there was a need to be addressed! This motivated them to think innovatively and seek ways to differentiate.

After introducing their solution to various HR managers in different-sized businesses, they got insightful feedback on what they really wanted to know and track within the workforce. It was clear that understanding their human capital could not be solved just by Watson’s Big 5 model.

Conducting more research, they came across two widely accepted alternatives in the scientific and business world:

1. The Myers-Briggs Indicator (MBTI) that measures:

  • Extraversion or Introversion
  • Preference for sensory information or analysis and interpretation
  • When making decisions: what is logical and consistent versus being dependent on the person and the context
  • Preference for structure or new information and options

2. The DISC Model that measures:

  • Extroversion or Introversion
  • Orientation to tasks or people

The duo found a public dataset in Kaggle with MBTI personalities which was sufficient to train their neural network. After incorporating their product to analyze the workers’ emails every hour, they realized that the data obtained varied depending on the recipient of the email and on whether the person was under stress. The aforementioned personality models are designed for quizzes, surveys, and personality tests that divide people into pre-established ‘types’.

"The existing models try to fit the universe in a box, instead of adapting the box to the universe."

While scanning through the whole Perlego library in search for books on personality theory and devouring all the scientific articles they could find, it appeared clear that no model was going to be sufficient. People are dynamic and change both with their internal (hunger, thirst, libido, pain, neurotransmitters, hormones, substances) and external context (friends, family, economy, trauma, culture). More importantly, the way we see ourselves (desired personality) differs from the way other people see us (apparent personality), and from the way we act (real personality).

Creation of our own Semantic Personality Analysis Theory

In the end, Erudit’s co-founders came to the conclusion that the widely accepted models were not exhaustive enough to be able to carry out a thorough personality analysis. It is important to treat them as dynamic belief systems and by no means as global absolutes. 

At the end we concluded that while we cannot dispense with models at all. We must handle them as dynamic belief systems and not as global absolutes. We decided to call our theory Semantic Personality Analysis due to its central hypothesis. It states that we are the beliefs that we are constantly repeating, and that we act upon ourselves, others, and the world according to that speech. Therefore, we do not try to enclose workers in a global category. We rather offer metrics on different aspects of their hourly mental states considering their historical behavior and cultural context.

“We decided to call our theory Semantic personality analysis due to its central hypothesis which states that we are the beliefs that we are constantly repeating, and that we act upon ourselves, others, and the world according to that speech.”
Ricardo Michel, CTO and Co-Founder of Erudit

Therefore, Erudit does not box employees into globally defined categories. On the contrary, Erudit offers metrics on different aspects of their hourly mental states, taking into account their historical behavior and cultural context.

We designed all sorts of scrapers to get texts from books, poems, songs, dialogues from movies and series, social media, forums, and hired a team of psychologists to tag them based on our theory and the metrics we knew were of interest to Human Resources Managers. At the same time, we designed an algorithm based on AI to create word vectors that encode their meaning from their co-occurrence in the same context, given that neural networks only work with numerical inputs and cannot be given the text without being previously processed.

We also designed other algorithms to remove the noise caused by common words, punctuation, links, etc. We created the architecture of our neural networks to express each metric in terms of risk, re-evaluated the results on new data and retrained the AI until we were satisfied with the result.

What has Erudit become?

Today, Erudit allows people to connect different communication tools (Slack, Google Workspace, and Microsoft 365) to the platform to allow our servers to obtain messages every hour–without storing any personal information, but rather aggregating and processing it–which, once collected, go through a series of neural networks that give metrics on:

  • The emotional state of the users (happiness, anger, sadness, fear)
  • Well-being (empathy, frustration, loneliness, self-esteem, irritability, resistance to stress)
  • Connection with their team, managers, and company
  • Calculated risk of anxiety
  • Calculated risk of quitting

As a result, these parameters help us build three key metrics called burnout risk, engagement, turnover risk levels. With the help of our technology, we hope to open the conversation about the importance of mental health in companies and to draw managers’ attention to teams in need of help and support.

Updated content: