Ethical Issues in Biomedical Text Mining

This chapter discusses the ethical issues that must be inspected before deploying automated or smart healthcare systems, as discussed in the book. In biomedical text mining, the nature of the data, such as text, is much more objective. However, the sensitivity of such data is likely to be hampered as and when the data is used within different infrastructures, such as in the medical, technical, and social domains. The automation of natural language processing (NLP) techniques for such sensitive data can fall short of ethical, particularly social and technical, regarding data collection, modelling, and self-learning models. The user’s agency needs to assure that the entire process during automation should ensure that ethics is followed. Informed consent for a patient’s data becomes a sensitive area of consideration. While the identification of anonymization has been used to deal with data privacy, it is imperative to apply an ethical framework carefully. These aim to ensure the context in which the systems will be used. Despite these data issues, critical questions persist over recommender systems, information extraction systems, and automation of biased data with large models. Biases can creep into the system as easily from scientific objectivity as from social problems and systemic injustice. Which model is being deployed, in what ethical framework, and in what domain is only the starting point to dealing with such data? Hence, it is important to inspect such systems for their data, design, and the social environment in which they could be used. These socio-technical considerations, which form a humanistic approach, are necessary to ensure the smooth functioning of such powerful technology.