The Science Behind Thinklytic

Sticky Post January 15, 2019 Joel Davies 0 Comments

Thinklytic combines the latest research in cognitive neuroscience, psychology, and data science to revolutionise the way that organisations make hiring decisions. Thinklytic was originally conceived by a number of top academic scientists at the University of New South Wales, Sydney – the no.1 ranked university for neuroscience and psychology in the southern hemisphere. These scientists combined two decades of research conducted by UNSW with the best research from hundreds of other university labs around the world. Thus, Thinklytic is the amalgamation of everything that our best scientists know about how to measure human potential and predict person-job fit.

The science behind Thinklytic can effectively be broken down into two key components: the measures and the prediction algorithms. As you’ll see, it is innovation in both these domains that makes Thinklytic so revolutionary.

The Measures

Over the last century, scientists have been studying how best to measure the human mind. For many years, scientists were able to make some headway with measuring basic traits like intelligence and personality type. However, many of these measures were blunt tools that lacked precision and were subject to considerable bias.

In recent years, breakthroughs have emerged from the field of cognitive neuroscience that have dramatically increased our ability to accurately measure the traits and abilities that make a person who they are. New breakthroughs in cognitive neuroscience are occurring almost daily. Much of this research involves the use of computerised tasks that participants complete while inside brain scanners. These computerised tasks have been refined over time to measure specific traits with incredible precision. The most accurate and relevant tasks from this research have been adapted into Thinklytic by adding gamification elements to make them as engaging as possible.

In addition to neuroscience-based games, Thinklytic also leverages the last research from the field of psychometrics to measure key traits and abilities that games can not yet measure reliably. For example, the Emotional Quizzes utilise a methodology known as the Situational Judgment Test to measure Emotional Intelligence. These were developed in line with best practices established by thousands of emotional intelligence studies that have adopted the same approach to measurement. 

Thinklytic measures 22 traits and abilities. Data from all of the games and quizzes is converted into percentile scores for each trait and abilities continuum. These percentile scores are calculated by comparing data on each continuum to data of a comparable population. Is this way, one can determine how the candidate compares to all other similar people who have completed the assessment previously. However, the Thinklytic platform collects much more than 22 sources of information. In fact, our platform captures over 400 different data points from every candidate to be included in our prediction algorithms.

The Prediction Algorithms

The Thinklytic prediction algorithms provide information about which candidates are most likely to be high performers in a particular role. For the last 50 years, traditional psychometric assessments (e.g. IQ tests and personality tests) have been validated by (1) administering the assessment to a large group of people in an organisation, (2) collecting job performance data for each person, and then (3) correlating scores on the assessment with job performance data at a later date. This approach has been published in many scientific papers and is the accepted method of determining whether an assessment predicts future performance. However, this approach has a number of very serious limitations, such as a built-in assumption that all job roles require the same traits, a reliance on a very small amount of data to make predictions, an assumption that a higher score on a measure is always better, and a failure to account for interactions between different measures.

In recent years, new breakthroughs in data science have emerged that allow the traditional approach prediction to be dramatically improved. A host of different machine learning algorithms have now been developed that significantly improve the prediction of job performance. This is achieved by accounting for non-linear trends, allowing much more information to be included in assessments and enabling much more complex interactions between data sources to be accounted for. Thinklytic uses a combination of these machine learning algorithms to predict job performance for specific roles. It does this by (1) administering the assessment to a large group of people in an organisation or job role (2) collecting job performance data for each person, and then (3) applying the machine learning algorithm to the data to predict job performance from the assessment data.

A unique property of the machine learning algorithms that Thinklytic employs is they allow prediction scores to be generated from new individual cases. This means that every person who completes the assessment can be assigned a suitability score based on how well their data matches high performers who have completed the assessment previously. Until recently, this was impossible. It means that a single score can accurately represent all the data from the assessment, significantly reducing the level of human interpretation required to make hiring decisions. 

The science behind Thinklytic is vast and comprehensive. It is based on the best research and technology that is available today. We are passionate about science and would be more than happy to answer any specific questions about the science underlying our platform. We’d love to hear from you. Just email us at hello@thinklytic.com to find out more.