As the latest guest for our department's Research Seminar Series, we welcomed Gui Liberali from the Erasmus University Rotterdam. He shared his recent work on improving Randomized Controlled Trials (RCTs).

RCTs are widely employed by researchers and practitioners to discern the most effective treatment option among multiple alternatives, such as optimizing advertisement designs. Typically, individuals (e.g., customers or website visitors) are randomly assigned to experimental groups upon entry with equal probability. However, this assignment strategy becomes suboptimal as soon as certain trials are identified to underperform compared to others.

The flaw in maintaining equal group assignment probabilities throughout the experiment is that it exposes some individuals to suboptimal treatments. Yet, allocating fewer individuals to the suboptimal treatment later in the experiment compromises the precision of treatment efficacy estimates. This dilemma creates a trade-off between maximizing earnings, achieved by assigning more individuals to the apparently more efficient treatment, and maximizing learning, achieved through equal distribution for more precise estimates.

Gui proposed an innovative procedure to enhance RCTs and manage this trade-off effectively. This approach dynamically adjusts group assignment probabilities based on the evolving information about treatment efficacy at each new entry point in the experiment. In his presentation, Gui did not illustrate the model with an example from Marketing, but with two instances from medicine where the above-mentioned trade-off transforms into an ethical dilemma. Earning vs. learning becomes saving lives vs. learning, which opened the door for discussion with the overarching question: Are RCTs ethical?

Thank you, Gui, for your visit and sharing your thought-provoking work with our community.