Ethics of Medical Archival Internet Research Data

Tsuriel Rashi, Elad Yom-TovJournal of Medical Internet Research 2023 · OpenReview

Abstract

Medical research based on internet archive data, which in some ways is quite different from other data-based studies, is becoming more and more common. Despite its uniqueness and the challenges that characterize it, clear ethical rules designed to guide practitioners in this field have not yet been written. This article points to the lacuna that exists in legal and ethical texts today and offers an ethically balancing alternative. Among other features, the balance is based on the famous three laws of robotics by Asimov and a series of values, including transparency, accountability, fairness, and privacy.

Introduction

As medical research increasingly moves from the clinic to the cloud, researchers are tapping into a goldmine of "internet data"—everything from search engine queries and social media posts to GPS pings. While this data has proven invaluable for the early detection of conditions like Parkinson’s disease, cancer, and stroke, it currently exists in an ethical "wild west." Traditional frameworks designed for clinical settings are often ill-equipped to handle the nuances of archival digital footprints.

In a recent viewpoint paper published in the Journal of Medical Internet Research, Tsuriel Rashi and Elad Yom-Tov argue that our current ethical frameworks are failing to keep pace with archival internet research. To bridge this gap, they propose a novel solution: adapting Isaac Asimov’s famous "Three Laws of Robotics" into a new ethical compass for data scientists.

The Ethical Lacuna in Archival Research

Traditional medical ethics are built on the foundation of informed consent and direct oversight by Institutional Review Boards (IRBs). However, archival internet data—data collected by service providers for non-research purposes—rarely fits this mold. The authors highlight several unique challenges that create a regulatory and ethical gap:

  • The Consent Gap: Traditional consent is specific, explicit, and voluntary. On the internet, consent is usually buried in "Terms of Use" agreements that fewer than 0.1% of users actually read. This "click-wrap" consent is a far cry from the informed participation required in medical trials.
  • Lack of Control: Unlike clinical trials where researchers design the environment, archival researchers have no control over how the data was originally collected or the "interventions" (such as user interface changes) that occurred during the collection period.
  • Regulatory Lag: While modern privacy laws like the GDPR and CCPA provide protections for personal data, they do not provide a comprehensive roadmap for the ethical use of that data to improve public health outcomes.

Methodology: Comparative Analysis of Ethical Codes

To understand the current landscape, the authors performed a comparative analysis of existing ethical codes from major global organizations, including the Association for Computing Machinery (ACM), the OECD, and various governmental bodies. They mapped these codes against core values such as transparency, accountability, and fairness.

Table 1: A comparison of values across various existing ethical codes, highlighting the gaps in current frameworks.

Finding that no single existing code fully addressed the nuances of medical archival research, they synthesized a new framework by adapting Asimov’s Laws of Robotics. This "Asimovian" approach was chosen because it prioritizes human safety and creates a clear hierarchy of obligations for powerful, autonomous-like systems—or, in this context, the automated data research processes that analyze millions of users simultaneously.

A New Compass: Asimov’s Laws for Data Research

The core output of the paper is the proposed Asimov’s Laws for Data Research, which establish a hierarchy of responsibility:

  1. Zeroth Law: Data research should strive to benefit humanity as a whole.
  2. First Law: Data research should take care to minimize injury to a human being and minimize harm to humans through inaction.
  3. Second Law: Data research must be regulated by human beings (except where such orders conflict with the First Law).
  4. Third Law: Data research must be protected (unless such protection conflicts with the First or Second Law).

By placing the "benefit to humanity" (Zeroth Law) and "minimizing injury" (First Law) at the top of the hierarchy, the authors provide a path for resolving ethical dilemmas where individual privacy might conflict with urgent public health needs.

Discussion: Beneficence and the Duty to Act

A key insight of this framework is the shift of "beneficence" to the societal level. Since individual consent is often impossible to obtain for massive archival datasets, the authors argue that the research must serve the collective good. This implies that findings should be published in the public domain to ensure they provide maximum societal value.

Furthermore, the authors introduce the concept of harm through inaction. Borrowing from Asimov’s First Law, they suggest that if a researcher identifies a significant public health risk within a dataset (e.g., a cluster of symptoms indicating an outbreak) but fails to report it due to bureaucratic or privacy concerns, they may be violating their primary ethical duty.

Finally, the paper warns that anonymity is not a panacea. Even with standard anonymization, "malicious researchers" can often de-anonymize users through data linkage. Therefore, the ethical responsibility for protection must lie with the data owners and researchers' conduct, rather than relying solely on technical algorithms.

Conclusion

The authors acknowledge that while these laws are presented as "absolute," they are intended as a heuristic guide for ethics committees and professional organizations. As AI becomes more integrated into data processing, the line between "data research" and "robotics" will continue to blur. For researchers in AI, medicine, and data science, this framework serves as a vital reminder that the power of big data comes with a profound responsibility to ensure it is used safely and for the greater good of humanity.

Summary updated

Abstract

Medical research based on internet archive data, which in some ways is quite different from other data-based studies, is becoming more and more common. Despite its uniqueness and the challenges that characterize it, clear ethical rules designed to guide practitioners in this field have not yet been written. This article points to the lacuna that exists in legal and ethical texts today and offers an ethically balancing alternative. Among other features, the balance is based on the famous three laws of robotics by Asimov and a series of values, including transparency, accountability, fairness, and privacy.J Med Internet Res 2023;25:e43754doi:10.2196/43754
BibTeX
@article{rashi2023ethics,
  author = {Tsuriel Rashi and Elad Yom-Tov},
  title = {{Ethics of Medical Archival Internet Research Data}},
  journal = {Journal of Medical Internet Research},
  year = {2023},
  url = {https://openreview.net/forum?id=bPefJRBKY8},
}
Figure