About two years ago, a group of highly talented senior researchers from a startup named Unanimous approached Dr. Matthew Lungren, assistant professor of pediatric radiology at Stanford University Medical Center, with an inquiry: in what ways, if any, could technology designed to harness the power of collective human intelligence benefit the world of radiology or medicine in general? A collaboration between these researchers and Dr. Lungren commenced soon after, around the time when Stanford researchers released data showing that the detection of pneumonia on x-ray could be accomplished by AI with the same level of accuracy as human radiologists. The investigative question then became whether or not collective human intelligence could outperform the independent power of both AI and human radiologists.

 

So, what exactly is collective or 'swarm' intelligence and how is it better than just having a conversation with colleagues about a particular problem or decision? Dr. Lungren describes it like this: "If you can imagine a puck on ice that can be slid around, and each person has a magnetic force that they can apply to that puck to pull it toward the answer they believe is correct…eventually a decision is reached…and it's fascinating to see how accurate they end up being as a group." Unlike sitting around a table with your colleagues and eventually coming to a conclusion, no one knows the identity of anyone else in the swarm, which immediately eliminates the hierarchical and sociological influences of decision-making processes that involve perceived leaders or people of power; even if subtle, the dynamics that emerge from such heterogeneous groups often play influential roles on the final decisions that are made. Swarm intelligence removes that influence, and replaces it with distributed anonymity in decision making.

 

The possible use cases of this technology extend far beyond the world of radiology and hold promise for a future filled with better, more accurate diagnoses and decision making in medicine, but that's not to say it's not without its challenges and drawbacks. Press play to hear the full conversation, learn more by visiting the web page of the ones who started it all (unanimous.ai) and feel free to email your questions to Dr. Lungren at mlungren@stanford.edu.