AI for decision-making
This research stream aims to investigate how AI can contribute to better decision-making now without losing track of the potential negative externalities algorithms can have.
In this research stream, we investigate how the many new technologies that originate(d) from the fields of machine learning and artificial intelligence can be leveraged to help decision-makers, institutions, and organisations better serve societies and its individuals right now. We do so without losing track of the potential negative externalities they might bring about (e.g. lack of interpretability, fairness, loss of information from human experience).
Recent work by e.g. Mullainathan et al. has shown that ML techniques can already greatly contribute to decision-making when applied in the right context. One of these contexts are called policy prediction problems which occur when a prediction of a future state of the world (e.g. the weather) is enough to make a decision without needing to know the exact causal mechanisms behind it (e.g. the value of bringing an umbrella).
Crucial dimensions in identifying the right contexts and implementations of current existing AI technologies in business settings are liveness vs security, interpretability and transparency, human-AI interaction, and distributional impact across the target population.