New developments in machine learning have the potential to dramatically improve experimentation and clinical trials by making them more efficient, safe, and accessible.

From a business perspective, the massive adoption of the Internet by virtually everyone made it easy to run large-scale online experiments to identify the best ways to communicate and interact with consumers through banners, website design, emails, promotions, product recommendations and others. However, for firms to succeed in their experiments they need reliable and credible machine learning and causal inference methods that allow them to trust the findings and results, understand the extent to which their findings can be expected outside of their sample, and separate causation from correlation. Our researchers have robust statistical and methodological training that helps firms extract the most from their experiments and online investments, improving online and offline metrics and KPIs such as conversion and bouncing.

From a clinical perspective, efficient experimental designs of trials of pharmaceutical drugs reduce the pressure on public and private health research funding, and minimize patient harm because they require smaller samples than RCTs, and allow for shorter trials. We develop and apply leading-edge clinical trial methods, such as multi-armed bandits and Bayesian adaptive designs, to help pharma companies and medical researchers with the statistical- and methodological transition from RCTs to adaptive trials.