What challenges do data scientists face

Why the future of data science needs to be holistic

My curiosity to understand the world around me has always driven me. It paved the way for me from studying theoretical physics to my job as head of the artificial intelligence (AI) research team at Lindera. And who knows, maybe without her I would have pursued my second great passion and would have embarked on a career as a rock musician as a bass guitarist. Instead, I followed the tradition of a science that has tried since the beginning of the 17th century and Galileo Galilei to explain the world and its relationships mathematically precisely. But here, too, my curiosity drove me and that's how I ended up where I can fathom new contexts: In the research and development department at the healthcare startup Lindera.

Here we develop AI solutions for the care and health industry. Because our society is aging and the number of vacant nursing positions is increasing. Digital health solutions can help us with these challenges of our society: be it through falling costs in medical technology, accelerated development times, the relief of nursing staff or the improvement of the quality of care. These efficient, digitally supported processes are based on innovative intelligent technologies and big data.

Our mobility analysis is an example of how easily the quality of care can be improved with apps, smartphones and artificial intelligence - for nurses, caring relatives and those at risk of falls. A 30 to 40-second video with a smartphone camera of the senior citizens' gait and a short psychosocial questionnaire are sufficient for a detailed analysis of the gait. Based on AI algorithms, the digital health solution provides an individual catalog of measures to prevent falls.

Decision Intelligence: A new discipline that helps us make decisions

However, it begins long before the development of digital health solutions. Because on what basis does the algorithm make its decisions? This is where the relatively new discipline of Decision Intelligence Engineering (DI) comes in. As an association from psychology, applied data, social, neuroscience and management science, it deals with all aspects of selection - or like Dr. Cassie Kozyrkov, Chief Decision Scientist at Google and one of the best-known representatives of the discipline, says: "Decision intelligence is the discipline of turning information into better actions at any scale." improve the world around us with the help of data. But how does it do it and why is it so important in the age of AI and big data?

Beyond data science

In our work as data scientists, we handle large amounts of data, which form the basis of our solutions. Ideally, the data scientist receives an order from the decision maker who knows exactly what he needs. To do this, the data scientist collects data, evaluates it and develops a machine learning system in order to implement the problem solution sought.

But ideal scenarios are too rare in the real world. It is not new to us that the decisions made by managers and the selection of data are by no means free from the subjective bias of their creators. Machine learning systems are only as intelligent as their training data. Each technology is thus a reflection of its creators and systems. The same applies, of course, to AI. In order to use AI responsibly, it is therefore essential to develop the right skills to uncover and understand decisions. Data science alone cannot cover this.

That is why we need DI. As a superstructure for economists, social scientists, neuropsychologists, teachers, managers and many more, it combines their specific strengths. The holistic approach analyzes the causal structures between factors and identifies the best measures to achieve a certain result. To achieve this, decision-makers need to be trained not only in the purely quantitative sciences like data science, but also understand how to use behavioral sciences to make data-driven decisions. That means making a decision effectively - even before you look at the data.

DI as a perspective for Lindera

At Lindera, we invest a lot of time building the right architecture for reliable and scalable decisions. DI represents the nervous system of our company. As data scientists, we do not brood over our data in isolation from the other areas. Rather, we are in constant contact with our nursing scientists, doctors and economists and thus jointly set the course for our development. This way we can avoid adopting mistakes or unconscious bias.

Our short-term goal is currently to further optimize the gait parameters for the case analysis and to validate them in studies. In the next step, an AI-based catalog of measures is created from which individual recommendations are derived. The aim is for the system to learn such individual adjustments itself. So far, this has been done on the basis of rigid, logical criteria. The big challenge for us is that we only have a small amount of data available. Traditional approaches to deep learning fail to achieve their goal. In case analysis, we are dealing with neural networks that are difficult to analyze. However, with the help of DI, we can delve deeper into the matter and understand the basis on which the AI ​​makes its decision.