Abraham Maslow famously stated, “if all you possess is a hammer, everything appears to be a nail.” 

With all the hype with regards to Artificial Intelligence (AI), it’s easy to perceive it as an all-encompassing utility that can be deployed to provide solutions to healthcare’s plethora of problems. As a matter of fact, there are various debates occurring over what AI and Machine Learning (ML) are and are not capable of achieving. The actuality is that there are an array of Artificial Intelligence technologies that provide solutions to disparate types of issues. There are unique use cases for every type of Artificial Intelligence, ranging from advanced analytics to deep learning to casual AI. However, as we start to tackle healthcare’s most complicated questions, we are required to comprehend which solutions – black box or white box solutions – furnish the correct answers. 

Black Box AI and Deep Learning 

There are several variants of machine learning, but the one that has produced the most discourse is deep learning. Deep learning is a variant of machine learning where neural networks go about learning from data and categorize patterns. With the passage of time, neural networks start identifying these patterns and essentially get smarter. 

Deep learning is one of the most debated forms of AI as it has proven to be really effective for various healthcare applications. Image recognition is likely the area of biggest impact as it is leveraged to diagnose early illness by identifying tumours and other anomalies usually invisible to the naked eye. 

Deep learning and neural nets are assisting us in going about learning patterns embedded in the data. Algorithms are developed that make predictions, but these forecasts are made in a “black box” meaning the underlying explanation or “the why” is not known. That does not make deep learning any less efficient for the issues it is attempting to identify solutions to. When leveraged to identify patterns and make correlations deep learning is a really efficient solution. 

Although, the fundamental restriction of deep learning is its incapability to go about learning the mechanism through which something occurs. This is especially problematic within the healthcare domain. With no learning the underlying biological mechanisms, it’s not possible to detect when to best interfere to influence patient health, to comprehend when one medication is better than another for a specific individual, or how an illness progresses. These questions are crucial when attempting to treat or cure illnesses or generate new therapies.  

How white box Artificial Intelligence influences healthcare 

So what is the precise meaning of white box AI? White Box AI is a classification of artificial intelligence that features transparency and is explainable. Casual machine learning or casual AI is a capable variant of white box Artificial Intelligence that leverages data to learn the underlying biological mechanisms and completely explain the outcomes. 

Unlike in correlation and predictive frameworks, the mathematics of casual inference and reasoning is able to evaluate – in parallel – a massive number of hypotheses to decide cause and effect relationships. It assists us to comprehend why individuals are becoming ill, what’s behind their illness, how to detect new treatments and how to go about applying the correct treatment to alter it. 

A good instance of black box vs. white box AI is centred on lung cancer. For several years, scientists have found it a challenge to establish a connection between smoking and lung cancer. They were well aware that was a strong correlation but what other factors contributed? And as the old saying goes, correlation does not always imply causation. Probably enhanced healthcare was better able to diagnose lung cancer. Or perhaps the influx of industrial pollution was behind the increase. Or probably it was due to a genetic disposition. It is worth keeping in mind that correlation is a measurement of how closely two things are related to each other but doesn’t establish the cause. It took decades and following hundreds of thousands of patients to establish proof that smoking caused cancer. 

Casual AI is capable of accelerating these variants of discoveries considerably. In The Book of Why, Dr. Judea Pearl utilized the theory of casual inference and leveraged casual machine learning to demonstrate how smoking is behind lung cancer. Leveraging Bayesian mathematics, Dr. Pearle generated a casual calculus consisted of powerful algebraic rules and leveraged them to make inferences with regards to casual relationships. Utilizing the capabilities of white box AI, Dr. Pearl was able to look at the several potential models with regards to the causes of lung cancer. By means of manipulation, he was able to remove other potential causes till smoking was the sole association that prevailed – therefore proving causality. The white box solution had therefore, come into prominence. 

Leveraging the correct tool for the correct problem 

Deep learning has achieved several amazing things and has played a central part in enhancing several facets of healthcare. Automating administrative procedures, streamlining outreach to patients, natural language processing, and increasing medication adherence have all been largely assisted by this variant of AI. 

However, as we move in the direction of personalized medicine – obtaining the correct treatments to the correct patients at the right time – we require the capability to detect and explain the cause and effect relationships underlying the biology of medicine, which needs white box AI. 

Comprehending why some patients have quicker progression in a particular illness, why a specific gene mutation means a treatment won’t function or why specific patients reacted to one therapy versus another, is a very different issue than detecting patterns within a diagnostic test. White Box AI is the next stage in the AI revolution. It furnishes to us the opportunity to know more about cause and effect relationships at scale, which is precisely what will enable us to truly cure illness and revolutionize healthcare.