Huge congratulations to STAI CDT graduate, Dr Mackenzie Jorgensen, on successfully defending her thesis, ‘Mitigating Negative Impacts from Socio-Technical AI’.
We spoke to Mackenzie about her PhD thesis, being part of STAI CDT and what her next steps have been.
What is your thesis about?
My PhD thesis is titled “Mitigating Negative Impacts from Socio-Technical AI.” The type of AI systems I study are classification systems that can be used in high-stakes decision-making contexts. By socio-technical AI, I am referring to systems that are affected by our society (e.g., bias training data due to systemic injustice and inequality) and those designing and deploying them, and they also have serious consequences on people’s lives (e.g., loan granting).
While there is a great deal of work on mitigating bias from classification systems, it is unclear if those mitigation methods actually protect historically disadvantaged groups from harmful or discriminatory system predictions.
My thesis shows how common bias mitigation methods, under certain conditions, can perpetuate or exacerbate harm to these groups. I conduct multiple experiments within the financial domain and provide different ways of quantifying impact. In addition, I propose cost-sensitive learning as an alternative bias mitigation method to better uplift historically disadvantaged groups. I also provide a categorization of different bias mitigation methods based on how they make classification systems reliant on demographic group variables which can have implications under discrimination law in the UK.
How have you found being part of the STAI CDT?
I have really enjoyed my time at the STAI CDT from all of the people I have met and with whom I have become friends. I also have appreciated the wonderful support from Liz as the director and my PhD supervisor; she does so much for us students and I am so grateful for her and her continued support.
What are your future plans now that you’ve passed your PhD viva?
I am a researcher at the Alan Turing Institute in London on the European Lighthouse on Secure and Safe AI (ELSA) project, where we are focused on human oversight and AI governance, and mitigating the harmful impacts of uncertainty and bias in AI-assistance contexts. We are conducting research into how human oversight and governance mechanisms are applied to deployed AI systems and into the legality of uncertainty-based algorithmic interventions applied to AI-assistance contexts.
Any advice for anyone still doing their PhD or thinking of starting one?
No matter how many PhD students you talk to, no one PhD experience is the same. It is incredibly self-driven, so you have the power to turn it into something that you will find most impactful.
We look forward to seeing your career progress outside the STAI CDT and are excited to hear more about your new role. Congratulations again Dr Mackenzie Jorgensen.