Shaping AI with Measurement
I think of AI system evaluations as measurements. The more precise the measurement, the more precisely we can align AI systems. I study how best to gather the measurements that we use to evaluate AI systems - often called the “gold standard” or the “ground truth” - and see it as the key challenge for the immediate future of AI. I bring in knowledge of the way other fields measure from my experience - VR-based field studies, hormones from hair and saliva samples, psychometric questionnaires, electro-cardiac measures using the VU-DAMS, in Psychology in Criminology, in academia and industry - to the measurement of AI behavior.
My PhD thesis aimed to understand how to design, collect, and analyze ground truth, in challenging situations where there is no single truth, but rather multiple perspectives. It also informs generally how to “ground” psychological phenomena in text. The case study of my thesis involved measuring the basic values in song lyrics and political speeches. I brought knowledge from the Social Sciences, particularly psychometrics, to inform how to go about the process of “grounding” as a Trustworthy Artificial Intelligence (AI) systems PhD candidate in the Intelligent Systems department at Delft University of Technology, in the lab of Cynthia C. S. Liem.
I’m now looking to build on my experiences in a way that benefits as many people as possible. Download my CV or connect with me on LinkedIn.
Education
PhD in Computer Science,
(ABD) Delft University of Technology
Research MSc in Psychology,
Vrije Universiteit Amsterdam, Cum Laude
BA in Philosphy and Political Science, CUNY Queens College