Algorithmic Decision Making
Taking the right action, directed at the right target at the right time constitutes some of the most impactful applications of machine learning. Actions, unlike predictions, have consequences and therefore in seeking to take the right action, one must seek to understand its causal effect.
Nathan Kallus, Cornell Tech/Operations Research and Information Engineering, is extracting causal-effect-maximizing personalized decision rules from observational data in sensitive applications. Observational data, plentiful in domains such as medicine and civics, lack experimental manipulation. Isolated causal effects are obscured by complex selection processes—a phenomenon known as confounding. Such data are also messy, noisy, biased, and often missing. Despite the promise of rich and plentiful observational data, current approaches cannot handle the unique challenges it poses. It, therefore, can lead to unreliable, unsafe, and unfair decision making unfit for sensitive applications.
Kallus is creating a comprehensive framework of rigorous theory and robust methodology to address this gap and enable trustworthy decision-making systems, trained on observational data. To do this, Kallus is developing methods and theory for algorithmic decision-making in the presence of unobserved confounders.
He is also developing methods and theory for robust and optimal weighting for policy learning to address issues of stability, limited overlap, time-to-event data, and noisy and missing observations. Kallus is investigating algorithmic fairness of decision policies, trained from observational data. Ultimately, the research will advance knowledge at the intersection of machine learning, causal inference, and optimization.