Community research

For this edition, we've invited Dr. Nikolaus Poechhacker, postdoctoral researcher at the research group Complex Social & Computational Systems of the IDea_Lab, University of Graz, to discuss the team's research on social media recommender systems.

You’re currently working on a project on social media algorithms for societal good. Can you share what motivated you to explore this topic?

I have been working on recommender systems from a social science perspective in my PhD, though on a much smaller scale. While it is clear that social media has many detrimental effects on our democratic societies, it is at the same time a tool that allows for exchange on an unprecedented level. Recommender algorithms are a central infrastructure for social media - and thus they are a means where we can positively influence the discourse that happens on social media.

What also fascinates me are the many moving parts that must be taken into consideration. How do we conceptualize the "societal good" for social media? And how do we translate normative questions of "societal good" into concrete recommender algorithms? So, on the one hand it’s the scientific puzzle, but much more also the question of how we can have a positive impact on our digital lives.

How are you approaching the challenge of discovering how such recommender algorithms would work?

We are a very interdisciplinary team that brings together expertise from complex systems, computer science, and the social sciences. In our work, each perspective has a distinctive but interconnected "task", so to say.

First, we try to understand normative ideas about good civic discourse derived from social theory and an empirical stakeholder process. Then, we will translate these insights into algorithmic designs for recommender systems and operationalizing the normative claims into metrics that can serve as objectives for the recommender. And lastly, to test the effects of these recommender algorithms, we will run them against simulations of different types of social media sites to see how they do. The challenge here is that the normative ideals, the algorithmic design, and the design of the simulations, constitute a complex system itself. This requires a constant exchange of the different perspectives, which will, hopefully, end up in a result that is more than the sum of its parts and leads to insights into how we can design recommender systems for societal good.

Where can people find more information about the project and your wider work?

The best place to start would be the website of the IDea_Lab of the University of Graz.  There you find information about the project and the team.

A recent (Open Access) publication that provides a good overview of the project can also be found here. And, you can reach out to us. We are always happy to have exchanges on the topic.