‘My outcomes are somewhat counterintuitive’

In his home country USA, in the state of Michigan, Spencer Poodiack-Parsons was struck by the financial ruin for many individuals in the infamous MiDAS algorithm case. Here, 93% of the welfare fraud charges, resulting from the use of the system, were later proven to be erroneous.

‘In Holland we are familiar with the Toeslagenaffaire,’ Spencer says. Pinpointing the issue: ‘Multiple countries face similar examples. The central theme in my PhD work is to track down how humans working with new technologies such as artificial intelligence in interdependent systems, ultimately determine the outcomes. As clearly as possible.’

Police

Where strong opinions and AI meet, strong confirmations may arise, leading to dangerous perceptions of certainty. An example that comes to mind is police acting in specific neighborhoods known for their high immigrant population rates. Spencer: ‘Many officials thought the majority of the inhabitants were cheating, leading to wrongful arrests and charges. These arrests can then feed algorithmic systems which further target the same populations.’

Optimist

Such observations are an important input for Spencer’s PhD research. ‘If we manage to use AI in a wise and human centered way instead, it may help police officers and other members of official authorities to act less biased,’ he believes. Here is the optimist in Spencer speaking, both as a person and as a researcher.

Since 2019 the sociology aspects of humans working with rapidly arising computer technology started. ‘Now it is the right time to really make impact,’ Spencer believes. ‘The blind optimism in the times of social media and internet emerging, forever is a thing of the past. AI, ChatGPT and other large scale system algorithms may be hyping up at the moment. However, we know better than ever how micro-level socio-technical interactions between human agents and algorithmic technologies, can impact macro-level behaviour in unexpected ways.’

Interdisciplinary

In his research Spencer takes an interdisciplinary approach, combining the tools of data science with the insights of public administration. Before starting the PhD project Spencer had a working career in a high-tech company in the USA. ‘And, I am informed by the scientific expertise of my supervisory team at the University of Twente,’ he adds.

First analyses

In his early stages of PhD research Spencer is working on theoretical concepts of constructing hypothetically ‘unbiased AI algorithms’. And then testing them in simulations. He is hoping to demonstrate how researchers must consider factors beyond the technical performance. Even hypothetically perfect AI can lead to risks if the changes to human behaviours are not understood.

Spencer: ‘At this stage of research I am focussing on the conditions of benefits and risks. These first analyses are necessary to create follow-up pathways, to then create real impact. The research being useful in practice is my ultimate goal. The expected results are somewhat counterintuitive, in the sense that I believe my research will recentre the humans involved. I aim to bring a more holistic view, studying the technology as part of the entire interdependent system alongside human beings and institutions.’

Would you like to tell something about your background?

I received a BA (bachelors) from the University of Connecticut in Political Science in 2014 and then an MS (masters) from the University of Notre Dame in Data Science in 2020 (both United States universities). I am 32 years old.

How far along are you in your PhD program?

It will be two years in September and then I have two more years to go after that.

Which faculty are you affiliated with?

I work in the Public Administration section within the Behavioural Management and Social Sciences (BMS) department at UT.

Are you working full-time on your PhD research?

I am a full-time PhD researcher.