51 Making MetaDataCount Computer Vision News decides what should be labelled, what definitions are used for the labels, and what is the relation between the data that is possible to be collected and the labels. They employed ethnographic methods involving three health-tech organizations in Northern and Western Europe, and interviewed 46 participants from East Africa, Northern and Western Europe, including medical professionals, data scientists, designers, management, and other relevant stakeholders. Through this process, they identified five key external and internal factors affecting the creation of medical imaging datasets. Three external factors include regulatory constraints, the context of creation and use, and commercial and operational pressures. Two internal factors comprise epistemic differences and the limits of labeling. They concluded emphasizing the importance of considering these factors when discussing what constitutes high-quality data. This consideration is crucial for ensuring responsible AI design. ② Annika Reinke was the speaker of the second talk. Annika is the Deputy Head and Group Lead Validation of Intelligent Systems at DKFZ, Germany.
RkJQdWJsaXNoZXIy NTc3NzU=