Beyond the fourth paradigm of science - big data

Over the last few years I’ve made a number of predictions that have come true. Most notably, against all polls and bookies, Donald Trump’s election victory, and Scott Morrison election victory of 2019. I made the headlines, and the name Nostradamus was thrown about with abandon. True, it may have seemed as if I was performing some dark conjuring act of predicting the future, but really it’s science, data science, and it’s going to change everything.

Data is everywhere. It’s all around us and it comes in many forms. A single piece of data, in and of itself, does not hold much meaning. But when data is combined with additional data it adds value to any information that can be extracted. Take, for example, a location displayed by an individual GPS device. By itself, it’s not of much value, however, when we combine that data with data from other GPS devices we can then extract a whole range of useful information, like what is the fastest route of travel or an estimate on travel duration. Knowing how to use data and the information extracted from data is increasingly important in every facet of the modern world.

So rather than magic, it comes down to data. When we have a sufficient volume and sufficient variety of data - and we know how to extract meaning from these data - predicting future events can be done with a high level of accuracy.

But how did we get here? If big data analytics is a scientific revolution, then we need to consider how the paradigm of science has evolved to reach this point.

At first, scientific research was based on observations and experimentations. This is the first paradigm - Experimental Science. Latter researchers attempted to create theories and hypotheses and models to help explain natural phenomena. This stage is the second paradigm - Theoretical Science.  In this stage an hypothesis (or theoretical model) is essential. This theoretical model has no pre-determined outcome. It has to be something that can be supported or disproved through experimentation or observation. These theoretical models can seem like predictions, because not only are they able to explain previously observed phenomena, they also seem to predict new phenomena yet to be observed.

Advances in computer science and the exponential increase in computing power enabled the third paradigm of science -  Computational Science. This paradigm relies on computational modeling and simulation. Computational simulation also involves the construction and manipulation of a model of some part of reality. As with theoretical science the goal is to understand some natural phenomena. The difference here, however,  is that computational science enables simulation of many models in a relatively short period.

So, science evolved from Experimental to Theoretical and then to Computational science. However, recent advances in data science and the availability of a vast volume of diverse data coupled with high-powered computing has enabled the fourth paradigm, Data-intensive Scientific Discovery - Big data analysis.

Welcome to Big Data Analytics 

Traditionally, with such massive volumes of data, it was difficult to gain any meaningful insight. To reduce complexity and gain some level of insight, the data would often be reduced by statistical processing. But now, with new big data analytic techniques (data science, AI) we are gaining insight into data without traditional reductions. And what we’re learning about the world is startling.

While my election predictions garnered a lot of attention they were just convenient testing grounds for methods and algorithms I’m developing. Nonetheless, being able to predict an election with decimal point like accuracy has boosted confidence in other projects I’m involved in. With these projects, in areas such Tourism, Environment and Health, I start to understand these industries in new ways.

In tourism, for example, I can extract the number of tourists visiting a specific area. I can see what their interests are, their trajectory of travel. I understand the reasons behind their travel and even how they feel about their travel while they’re doing it. In terms of the environment, my work helps me predict environmental changes in ecosystems. Using the multitude of underwater videos posted on social media from holiday makers on the Great Barrier Reef I can see the changes taking place underwater over time and where those changes are heading. Other big data projects I’m involved in can help understand the patterns of traffic congestion, and how freight transport can be more efficient. And still others that can help monitor community wellness, identify causes of major depression, and pinpoint factors that influence obesity. The list goes on and on. Big data sheds light on almost every facet of modern life.

What is beyond the fourth paradigm of science?

The future will be one of even greater accuracy of data informed predictions. But to achieve this, new methods need to be developed that can dynamically learn from data. Not only will these new methods need to learn, they will also have to adapt to changes as they occur, as well as predict possible future changes and then, in turn, consider each of them. This would be achievable by developing new machine learning methods, beyond active learning, which would take into consideration (in addition to existing knowledge such as Wikipedia) vast volumes of diverse data, arriving at high velocity, while at the same time taking into consideration Veracity (how much can we trust the data?). Such human-like reasoning and intelligence will be able to create a theoretically infinite number of hypotheses and prove or disprove them in real-time. And when this happens we’ll find ourselves in the fifth paradigm of Science  - Infinite Science.

Bela Stantic is a Professor in Data Science and founder and Director of “Big Data and Smart Analytics” Lab at Griffith University. Professor Stantic is internationally recognised in the field of data analytics and efficient management of complex data structures. He has successfully applied his research to tourism, the environment, as well as to the health domain. He has published more than 160 journals and conference peer-reviewed publications. Professor Stantic made media headlines and showcased the power of Big data analytics with his correct predictions of the USA election, Brexit, the UK election, and particularly the last Federal election. For his help in modelling and predictions during the pandemic, he was named "Public Health Champion".

To visit Bela's Griffith Experts page click here