C2-g is a small data-driven consulting company with the mission to improve human decision-making under risk and uncertainty. Our approach utilizes best practices from computational cognitive science, machine learning and artificial intelligence to quantify human decision-making. Quantifying human performance allows for us to make an impact across several areas important to government and industry. We offer our services on a pro bono basis, so feel free to contact us with questions about how we can help address your data-driven needs.
Erik J Schlicht, PhD, is the founder of the Computational Cognition Group, LLC. He conducted research at Harvard University, MIT, Caltech and the University of Minnesota; his research has also been covered by several media outlets. Visit his professional web page for further information about his research experience and a sample of his publications.
Decision-making under uncertainty is critical to many domains, but humans are notoriously irrational when making decisions. In part, this is due to the fact that humans exhibit cognitive biases that adversely impact the quality of our decisions. Quantifying decision-making allows for these biases to be identified and mitigated by providing appropriate information. Decision support technology can be used to provide such information, but it necessitates that information is dynamically adjusted based on both its reliability and relevance to the current context. Fortunately, advances in computational cognitive science and machine learning allow for such an understanding to be obtained, provided the appropriate data exists or can be produced.
When new technologies are proposed, it is often unclear how changes to the system will impact safety, relative to the current level. Fortunately, using techniques from machine learning, we often are able to compute the relative risk in order to evaluate differences in safety between the systems of interest. In fact, our staff has experience using this approach in order evaluate the safety associated with both transportation and aerospace technology. For an example project, you may download a paper from this effort.
When a policy change is incorporated into practice, data driven decision-making requires that quantitative evidence be provided to the government in order to evaluate the program's effectiveness. Machine learning offers new statistical tools that can be used to derive valid quantitative metrics to evaluate a program of interest, or provide business insight. Our staff can tailor these metrics specifically to the questions you need answered. Moreover, we can train your current staff on how to compute these metrics going-forward.
Although artificial intelligence algorithms perform better than humans at decision-making, there are still tasks in which humans can easily out-perform even the best available algorithms. In these tasks (e.g., sensorimotor control), algorithmic performance may benefit from insight into the computational 'tricks' used by our brain to achieve superior performance. Our staff has successfully designed such biologically-inspired algorithms to improve performance of computer vision systems. Moreover, we have extensive experience in human computational sensorimotor control that can be leveraged in an attempt to improve robotic reach and grasp.
02.2017: C2-g released a paper on methods for using multifidelity simulation to estimate the risk associated with transportation technology