The U.S. Defense Advanced Research Projects Agency (DARPA) has launched a new program to solicit paradigm-shifting research ideas to revolutionize how scientists predict collective human behavior.
The program, known as Methodological Advancements for Generalizable Insights into Complex Systems (MAGICS), aims to address the problem that despite the rise of big data and machine learning, we’re still surprisingly bad at forecasting how large, dynamic human systems respond to change.
“For the past decade or more, there has been an assumption and hope that the explosion of digital data streams (e.g., social media, purchase patterns, traffic dynamics, etc.) combined with powerful machine learning tools would usher in a new era of research in complex, dynamic, evolving systems,” a DARPA solicitation notice writes. “[However] Despite many attempts, results have failed to meet expectations.”
The MAGICS opportunity, announced through DARPA’s Defense Sciences Office, invites individual researchers to propose innovative concepts that could form the foundation for a new science of social prediction.
As DARPA notes, today’s best statistical tools often falter when applied to real-world, evolving systems—whether it’s understanding how economies adapt to disruption, how populations shift under demographic pressure, or how societies react to technological upheaval.
At the heart of the MAGICS effort is answering the question: Can we develop new ways to model collective human behavior that outperform current statistical approaches and capture the dynamics of complex, evolving systems?
The Pentagon brain trust is looking for fresh frameworks beyond what’s possible with today’s machine learning models, namely systems that can handle the messy, recursive, and often unpredictable nature of human systems.
The stakes are high for national security. From forecasting the spread of misinformation to anticipating societal responses to crises, the ability to model human behavior accurately could offer profound advantages.
Yet DARPA acknowledges that researchers must overcome foundational challenges that large datasets and artificial intelligence have failed to address before these benefits can be realized.
Hellbound and Down 