Developed in close cooperation with data analysts from the City of Utrecht, DEDA is a tool-kit facilitating initial brainstorming sessions to map ethical issues in data projects, documenting the deliberation process and furthering accountability towards the various stakeholders and the public. DEDA helps data analysts, project managers and policy makers to recognize ethical issues in data projects, data management and data policies.

During this workshop, we will work with DEDA to discuss ethical pitfalls of an example case. You will get to learn how to work with DEDA so you can apply it within your own organization.

Given by Iris Muis (UU).

As the adoption and application of data science and AI is growing, ethical aspects are coming more and more to the forefront. What are some of the common ethical principles for responsible and trustworthy use of AI, and what progress has been made with AI policy and lawmaking globally? What are some examples of cases where things have gone wrong, and how easy or hard is it to prevent issues from happening. And as the pendulum is shifting from ethics to regulation, will it shift back to ethics again at some point? In this talk I will use examples of how things get interesting (and hard) when the rubber hits the road.

Peter van der Putten (assistant professor at LIACS, Leiden University & Director AI Lab, Pegasystems).

With machine learning methods becoming widely adopted in several fields, the need for data collection and synthetic data generation has experienced a similarly rapid growth. Massive datasets, new forms of data collection, and non-standard data types require quality assurance mechanisms that are able to keep up with and respond to these developments. A human-centric approach to data quality addresses these aspects by tying the societal context of machine learning deployment and data analysis to the outcomes of these processes. In this session, and drawing from experiences in several European research projects, I will explore some of the methods and rationales through which a human-centric approach can be embedded in traditional data science and machine learning pipelines.

João Gonçalves (assistant professor at Department of Media & Communication, ESHCC, Erasmus University Rotterdam).

Credit risk models are built with the primary goal of eliminating risky borrowers. However, during this process some customers might be deprived of credit not because they are risky, but due to the biases in herent in these models. In this talk we want to explore how to bring an objective approach to assess fairness which is quite subjective in its nature.

Daniel Merino Jimenéz & Özgür Çetiner (ING)

As machine learning becomes increasingly integrated into organizations, swift adaptations by criminals pose a challenge. In response to the emerging trend of leveraging machine learning for societal disruption, law enforcement must proactively stay abreast. Numerous machine learning projects are underway within police departments, prompting an examination of the consequences in terms of transparency, privacy, and accountability.

This session aims to delve into these implications, particularly in the context of anticipating large-scale protests from a law enforcement standpoint. Together, we will assess the possibilities and constraints through technical, juridical, and ethical lenses, shedding light on the multifaceted dynamics of machine learning implementation in policing.

Laurens Müter (data scientist at the Dutch police and PhD candidate at Utrecht University, Information and computing sciences)