Published: May 24, 2018
Forough Poursabzi-Sangdeh photoI’m Forough Poursabzi-Sangdeh, a PhD candidate in the computer science department. The goal of my research is to open up the black box and bridge the gap between humans and machine learning models by designing and empirically evaluating models and systems with humans in the loop.
 
Machine learning is ubiquitous, and continuous advancements have been made to improve the performance of these techniques. However, when it comes to deploying them in the real world, hesitation is common among practitioners, especially in critical areas such as health care and criminal justice. This skepticism is caused by a gap between model developers and end users.
 
Although machine learning has been extremely successful in automatizing many tasks and decision-making procedures, human involvement is still necessary for several aspects: humans should annotate and collect the data that is required for models, humans should develop, tune, and debug models, and humans should evaluate models and trust them. Despite the inevitable involvement of humans, their behavior is rarely studied and their goals are rarely considered in the community. In fact, these models are often considered black boxes, which makes human involvement complicated.
 
Part of the reason for the gap between model developers and end users is the difference in expertise as well as expectations. While the developers and designers of machine learning models are usually experts in computer science and statistics, the end users are usually domain experts (e.g., doctors, lawyers, social scientists). Domain experts need to trust these models before deploying them, while human trust and goals are not purely computational and, thus, are not a direct factor in the traditional machine learning design and evaluation pipelines.
 
I visited Boulder and CU several times before applying to the graduate school. I met great students and professors both in terms of the fantastic research they do and the care they give to others. I also loved the energy people have both on and off campus. My favorite experience so far has been meeting passionate freshmen, feeling the energy they bring in, and talking to them about my research.