I am a pediatric cardiologist and have cared for children with heart disease for the past three decades. In addition, I have an educational background in business and finance as well as healthcare administration and global health – I gained a Masters Degree in Public Health from UCLA and taught Global Health there after I completed the program.
“What people call AI is no more than using correlation to find answers to questions we know to ask. Real AI has awareness of causality, leading to answering questions we haven’t dreamed of yet.”
Tom Golway, information technology executive
Causal inference is the intellectual discipline that accommodates researchers to use data to draw causal conclusions by determining the independent effect of an element in a system (“cause and effect”). In our daily lives, we use causal inference to navigate the world as we face situations and solve problems. Many disciplines have an inherent interest in the concept of causality: economists, epidemiologists, sociologists, researchers, statisticians, and data scientists. Frameworks for causal inference include the Rubin causal model (also known as the potential outcomes framework) and structural equation modeling.
In healthcare, most studies aim to answer causal rather than associative questions. The classic manner in which causality is determined in clinical medicine and healthcare is the randomized controlled trial (RCT). Ethical and practical limitations, however, sometimes affect the deployment of an RCT. With the excitement about machine and deep learning in healthcare, a relative weakness of machine and deep learning methodologies in healthcare is that these tools, as good as they can be, usually do not provide causal inferences as these methodologies provide instead, correlations (correlation does not imply causation). The phenomenon of two variables having a high degree of correlation is termed multicollinearity. The prediction of an alternate reality is called counterfactual inference. Similar to its alternate inference (causal inference), counterfactual inference also has its own frameworks such as generative and Bayesian modeling. Counterfactuals are essential for causality as we often cannot test more than one element.
Machine and deep learning have yielded impressive dividends thus far, but cannot be the sole paradigm for solving problems. The future of AI, especially as we aim to reach artificial general intelligence (AGI), will need to learn similarly to how a child learns: understand casual relationships without overabundant data.
We are very excited to welcome you to attend in-person the AIMed Global Summit taking place January 18th-20th, 2022, at the sublime Ritz-Carlton resort in Laguna Niguel, southern California. This summit promises to be the most exciting yet, with Drs. Eric Topol and Daniel Kraft among the keynote speakers. We are all very much looking forward to seeing and learning from each other in person for human-to-human conversations and networking at this event. Book your place now.