黑料福利网

Understanding ML

Machine learning models are increasingly used to solve complex problems, from predicting outcomes to detecting anomalies. While some models, like ChatGPT or AlphaFold, attract public attention, many others quietly shape decisions in hospitals, universities, and government offices. Streppel鈥檚 research shows that these systems challenge traditional philosophical ideas about scientific understanding and raise new questions about how we interpret the world through data.

Explaining Decisions

The rise of 鈥渂lack box鈥� models has sparked debate about explainability. Streppel argues that explanation is not a fixed concept, but a contested one. Even transparent models require meaningful interpretation for doctors, judges, and citizens. Her work suggests that instead of defining explanation narrowly, we should focus on making information accessible and relevant to those affected.

Rights and Protection

European citizens have the right to an explanation when algorithms make decisions about them. But Streppel points out that this right often requires individuals to take initiative, which is unrealistic. Just as consumers can鈥檛 be expected to investigate every food label, citizens shouldn鈥檛 bear the burden of decoding algorithmic logic. Her research advocates for placing responsibility on developers and users of ML systems to provide clear, actionable information.

Embedded Values

Machine learning systems are not neutral. The data they use and the goals they pursue reflect human choices. Streppel emphasizes that these choices carry assumptions and norms, which shape societal outcomes. Her work calls for critical reflection on which values are legitimate and how they influence the design and deployment of ML technologies.

Yeji Streppel defended her thesis on October 30, 2025. Title of the thesis: '' . Supervisors: Philip Nickel and Emily Sullivan.