A philosopher’s unusual take: mathematical representations of diversity

Sina Fazelpour, a Ph.D. candidate of Philosophy at the University of British Columbia (UBC), spoke about his research and his unusual path to philosophy.

Before coming to UBC’s Ph.D. program in Philosophy, Fazelpour studied biomedical modeling and medical biophysics. Currently a philosopher, he works on philosophy of science where his previous training in engineering has carved out niche research projects that incorporate computer simulation and theoretical modeling into philosophical inquiries. By creating algorithmic-based models, Fazelpour examines qualitative ideas such as fairness, justice and discrimination that we commonly find in legal and political spheres.

Counterfactual reasoning and accountability

Sina Fazelpour.

One of Fazelpour’s interests is in counterfactual reasoning. Simply put, counterfactual reasoning entails “what if” or “if only” thoughts that we all entertain in everyday life to gauge the possibilities of how certain events can unfold, or what might have happened differently. We might think to ourselves: If only I had ordered another drink, it might have tasted better. Or we might question: What if somebody else were elected president? How could things have been handled differently?

“We use them [counterfactual reasoning] when we want to attribute responsibility,” Fazelpour said. In cases of discrimination, creating a reliable framework can help us better determine what factors, or what people, are responsible for undesirable outcomes in decision making, such as gender discrimination in the hiring process. It is especially important to make sure that we understand these causal relations correctly and that we base our judgment on good reasoning. If we are holding people accountable, our ideas should not be based on wishful thinking and random thoughts, Fazelpour commented.

One way we can apply counterfactual reasoning is to use computer simulation to test out how things might go differently given different parameters in a specific setting.

Diversity in statistical simulation

Statistical tools used in a designed simulation – or what Fazelpour explained as “a mathematical representation of the real world” – has the “allure of objectivity,” which allows us to run experiments in a controlled environment. Unlike actual settings, controlled conditions in a computer simulation can get rid of other confounding variables (such as interpersonal relationships, different grouping classes, etc.) and is useful for the purpose of studying phenomena relevant to the questions a study sets out to answer. In a medical case, for example, a simulated model can more accurately analyze whether an outcome results from the efficacy of a treatment rather than other random factors.

As another example, supposed we want to study the effect of cognitive diversity. Fazelpour explained how we can model one group of artificial agents that have the same way of thinking and learning (homogenous setting) versus another group of agents with different approaches (heterogenous setting). In that designed simulation, which is built on one specific aspect of diversity we want to study, we can then experiment and observe how things unfold.

Is a homogenous setting or a heterogenous setting more conducive to decision making? Fazelpour observes that in homogenous groups, people tend to have normative expectations that they must agree with each other; hence, they are too trusting and are not diligent enough in the way they process information. “Adding a little bit of diversity creates tension in groups but that tension can be productive,” Fazelpour said. A group of people with diverse identities can be useful because “people may be willing to share dissenting opinions” and “listen to more novel views.”

Consider the case of a jury, Fazelpour added on a similar note. “Is a homogeneous group where there is no friction and everybody listens to everybody else a good setting for coming up with good decisions in a criminal trial? Or is it better to have a diverse group? Of course from the perspective of justice and fairness, it’s better to have more diverse groups.”

This October, Sina Fazelpour will be starting a postdoctoral position at Carnegie Mellon University, where he will work on algorithmic-based decision making and its ethical implications. To learn more about his work and upcoming events, please visit www.sinafazelpour.com.