Bias, Machine Learning, and Liability—Can EU Policy Improve the ...
This chapter delves into the policy and regulatory challenges posed by artificial intelligence (AI) within the lifecycle of AI-powered products, considering past incidents of AI bias. It assesses how legal ramifications resulting from AI bias can be managed in scenarios involving explainable and non-explainable algorithms. The analysis extends to the developments in EU policy, particularly those geared towards crafting a holistic policy response suitable for an 'algorithmic society'. The chapter underscores the intricate choices faced by European policy-makers. On one hand, there is a need to establish adequate controls for AI-related risks, while on the other, it is imperative to ensure the competitiveness of European industries and societies.
Explainability in Algorithms
Explainability is a crucial requisite aimed at mitigating the 'black box' issue associated with algorithms. The concept of a 'black box' gained prominence following the US credit bureaus scandal. For further insights on this matter, refer to Pasquale's work from 2015. Additionally, research conducted by Balayn, Lofi, and Houben in 2021 delves into managing bias and unfairness in data for decision support, offering a survey of machine learning and data engineering approaches to spot and alleviate bias and unfairness in data management and analytics systems. An article detailing their findings can be accessed here.
Additional References and Resources
For further exploration of bias-related topics, the following resources are recommended:
- BBC's coverage of Google's apology for a photos app's racist blunder in 2015: Link
- An insightful paper by Eastman and Rix in 2022 on bias in expert witness practice, providing insights on sources, pathways of expression, and methods of reducing bias. The article is accessible here.
- A glossary of human-centric artificial intelligence by Estévez Almenzar, Fernández Llorca, Gomez, and others, published by the European Commission's Joint Research Centre in 2022. The glossary can be viewed here.
- European Commission documents outlining strategies and approaches towards artificial intelligence and data governance:
- Artificial intelligence for Europe (2018): Google Scholar
- White paper on artificial intelligence—a European approach to excellence and trust (2020): Google Scholar
- A European strategy for data (2020): Google Scholar
