TURNING HUMAN BEHAVIOUR INTO PREDICTIVE DATA

ZISHI Adaptive is an arm of the ZISHI business that augments human capability by fusing AI, machine learning and data science to uncover valuable, performance-enhancing insights previously invisible to see

The ZISHI Adaptive team includes some of the world’s brightest minds. It is a collective of highly respected trading experts, academics and data scientists who have amassed decades of experience and lead the way in their specific sectors.

In our endeavour to understand our trader’s behaviour and performance in more depth than ever before, ZISHI Adaptive has scrutinised over 45.5 billion data points. For the first time, we have been able to blend cognitive behaviour, personality traits and performance data. This has uncovered previously unseen insights to help de-risk our recruitment process and identify suitable candidates for trading roles.

Built initially as an internal recruitment tool, we have discovered ZISHI Adaptive’s capabilities aren’t limited to just our own use. At the very least, we can use our 45.5 billion data data points to predict performance for any likeminded business. But the unique ability of this tool also allows us to go so much deeper, combing our own data with a third parties unique data to uncover personalised empirical insights and predictions.

The next exciting part of our ZISHI Adaptive journey is the development of a trader mentoring tool. The tool has been designed to increase the support and likelihood of traders succeeding when they are trading live markets.

The definitive answer is NO. ZISHI Adaptive firmly believes that AI should be used to augment human capability, not replace it. We have endeavoured to maintain a ‘human in the loop’ approach at all stages of the development and deployment of our AI assisted tools. Our belief and recommendation is that the predictions are used as an aid to help maximise performance in combination with human intuition. The ‘human in the loop’ will always make the final decision.

Our ‘human in the loop’ approach helps manage this, along with the fact that our systems purposefully do not currently learn in real-time. Offline learning helps avoid some risks and unpredictable systems behavior that may lead to negative consequences of the system’s decisions. We also make sure we collect a holistic, rich image of data proxies. We put significant emphasis on rich and multimodal data collection, both automatically collected from a wide range of digital systems, audio and reports, and from designed data collection such as interviews and questionnaires. This ensures that our model of trading behaviour, performance and personal attributes is as holistic as possible.

The approaches we take are purposefully designed to be as transparent as possible for users of our technology. We safeguard our users by making sure their decision-making process is supported by other data-driven artefacts rather than predictions. When predictions are necessary, we make sure the users are provided with the metrics of confidence needed to understand the extent to which the predictions should be taken into consideration.

Our human experts have been at, and continue to be at, the core of our AI system design process. Their input and involvement has been split into three main areas: (a) They are co-designing the system, and are full collaborators in all design and implementation phases; (b) In the scope of the development of the machine learning models specifically, our experts’ domain knowledge and logic is foot printed in the computation of our features, as well as in crucial decisions such as assigning weights to features, defining similarity measurements and more; and (3) The validation of our models by integrating qualitative data collected from trader interviews with quantitative data collected from questionnaires and digital systems.