このコースは修士号の正式な単位として提供された最初のコースであるため、私は最初のコースとしてデータ倫理を選択しました。このコースは他のコースとは少し異なります。技術的な側面ではなく、データサイエンスの倫理的な側面を扱います。従って、数学やコンピューター言語に関する知識は必要ありません。但し、このコースでは、データ倫理に関する多くのドキュメントを読み、課題を書く必要があります。一般的に、ライティングの課題では、あるドキュメントに関する内容の要約、内容の倫理的問題、及び評論を含めて、500〜800語の記載が要求されます。

このコースでは、倫理原則の基本的な概念及びこれらの原則のデータサイエンスへの応用について学習します。これらの原則は、Utilitarianism、Kantianism / Deontology及び Virtue Ethicsです。簡単に言えば、Utilitarianismは結果に、Kantianism / Deontologyは道徳に、Virtue Ethicsは善悪に焦点を当てています。

このコースの課題のサンプルライティングをお見せしたいと思います。

Reference Information

Artificial Intelligence and Product Liability Law: Predicting Liability Risks Based on the Existing Regulatory and Legal Framework

Joe Fornadel and Wes Moran

September 11, 2020

JDSUPRA

https://www.jdsupra.com/legalnews/artificial-intelligence-and-product-24441/

Summary of the Content

These days the number of medical devices having artificial intelligence (AI) is rapidly growing, and many healthcare algorithms have been granted market approval from the U.S. Food and Drug Administration, particularly in cardiology and radiology. Most medical AI can help healthcare professionals with their medical practice, such as medical diagnosis. However, performance by medical AI is never perfect, thus it is disputed how to allocate liabilities for medical errors associated with medical AI.

Algorithms in medical AI are generally produced based on medical data set, and then will be applied to new data that are obtained after its market approval. These algorithms are classified into “locked” or “adaptive”: the former means unchanged algorithms despite new input of new data and the latter means algorithms changed over time based on results of new data. The former is simpler and easier to regulate, and the latter is more complex and difficult to regulate.

Issue(s) in Data Ethics

This article closely relates to the right to receive correct medical services. Performance by medical AI is dramatically improving and AI sometimes provides better results than certified specialists in the medical fields. However, AI also makes a mistake at a certain rate, and patients may suffer serious adverse effects. Therefore, it must be determined in advance how to deal with medical errors in association with medical AI. Traditionally, physicians would be liable for medical negligence. However, the manufacturers and the sellers may also take joint responsibility for such medical errors.

My Assessment

I take the view that the article sufficiently addresses the ethical issue. Firstly, the manufacturers and the sellers of the medical device with medical AI might owe product liability because of its design defects or failure to warn.

Regarding design defects, under the risk-utility test, a product has design defects “when the risks inherent in the design outweigh the benefits.” It is likely that the patient will have to prove the gravity of danger or magnitude of harm caused by the design defect as well as possibility and cost for a safer alternative design. It looks extremely difficult for patients to prove how the medical error by AI caused the serious adverse effects on them and how to design safer algorithms for the medical AI. Under the failure-to-warn theory, the manufacturers and the sellers have a duty to provide adequate warnings and instructions to the physicians who diagnose or treat the patients to avoid adverse effects by the medical AI.

Secondly, the manufacturers and the sellers might be liable under the theory of negligence if they knew or should have known the adverse results on the patients. However, the burden of proof is imposed on the patients' side.

Thirdly, the manufacturers and the sellers might be liable under the theory of breach of warranty. However, the patients have to purchase the medical device from the seller directly for its application.

I take the view that this issue must be resolved at the international level because it is possible to manufacture medical AI in a foreign country and import it to the U.S. This is different from traditional medical negligence cases.

According to Utilitarianism, medical errors by AI must be prevented as much as possible to maximize the happiness of patients in the country. Kantianism /Deontology and Virtue Ethics also morally requires reducing medical errors by AI.