Governance of High-Risk AI Systems in Healthcare and Credit Scoring

Abstract

MotivationGovernance of artificial intelligence (AI) systems is an evolving field focused on establishing the frameworks, policies, and practices necessary to ensure the responsible use of AI systems across various sectors (Ma ¨ntyma ¨ki et al. 2022).The need for AI governance stems from growing awareness of the significant challenges these technologies pose.Research and real-world examples highlight that AI systems can lead to unintended consequences, including biases, discrimination, inaccuracies, and opaque decision-making processes (Schneider et al. 2023).This raises questions in terms of algorithmic accountability, as it is necessary to determine who holds responsibility for algorithmic actions when systems operate in a harmful or unethical manner (Horneber and Laumer 2023).Algorithmic accountability involves identifying the parties liable for design flaws, operational errors, or misuse of AI systems, while also ensuring that mechanisms are in place to correct biases, address inaccuracies, and enforce transparency in decision-making.Defining these accountabilities is an important task in a governance framework for AI as many parties in the organization will be involved in setting up and running AI systems (e.g., data provision, decision on data sources, defining policies).Furthermore, these and other examples also show that AI systems can have far-reaching consequences for work, labor processes, employees (in terms of, e.g., technology acceptance, productivity, autonomy, identity) and organizations (e.g., structures, culture, leadership) as a whole.They therefore always (should) raise questions about the humane shape of work and organization in the context of digital governance (e.g., Bankins et al. 2024;Jungta ¨ubl et al. 2024).While governance and accountability of AI systems are essential, there are specific scenarios where they become

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 14 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.