Subject: Proposed Guidelines for AI

Subject: Proposed Guidelines for AI

Intelligence/Machine Learning Applications Guidelines for the Philippine College of Radiology (proposed draft as of June 23, 2018)
Regarding this proposed draft:
–This draft is loosely based on the GOOGLE AI principles found at https://www.blog.google/technology/ai/ai-principles/ which was posted by Sundar Pichai (google CEO) on June 7, 2018.
–This is only a guide and not an absolute in the development of AI/ML applications under the Philippine College of Radiology (PCR). It should be regarded as a living document that is willing to change and adapt based on the needs of the people and the demands of the time. –Erwin John T. Carpio,MD (Radiologist, FrontEnd Web developer, AI enthusiast).
Artificial Intelligence/Machine Learning Applications for Radiology should:
Be medically beneficial
The role of artificial intelligence and machine learning in the practice of medicine should be used to uphold if not improve the standard of health care.
It should not be used simply as a means of cutting down cost
Its purpose should not be to replace or compete with bonafide well trained radiologists but rather to help radiologists and clinicians in the practice of medicine so as to help improve the lives of patients.
Avoid creating or reinforcing unfair bias in the practice of medicine.
Algorithms and data sets can either reduce or propagate unfair bias.
Biases regarding race, ethnicity, gender, nationality, income, sexual orientation, ability, political and religious beliefs should be minimized unless these are medically and scientifically proven sound points of interest for health care (and not for financial, economic or personal gain).
Be built and tested for safety.
AI is new and the means of conducting safe AI/ML development is still in its infancy. However, awareness that such a concept of safety when developing AI/ML applications should be made known.
Be accountable to People (Doctors, Patients and Health Care Personnel/Practitioners)
AI/ML systems should not be the new masters of health care. Instead, development of AI/ML applications should allow for feedback, relevant explanations and appeal from all persons involved in its use and development.
Incorporate Privacy Design Principles:
This is subject to the Data Privacy Act of the Philippines.
Uphold high standards of scientific excellence:
To diversify AI/ML applications amongst the various fields of medicine where applicable by working with a range of stakeholders so as to compliment the needs of fellow clinicians and patients (not the needs of financial corporations).
Artificial Intelligence/Machine Learning Applications should:
NOT Cause overall harm. If there is unavoidable harm, just as in any practice of medicine, the risk-benefit ratio should be properly evaluated by a trained physician/radiologist and coursed/explained adequately though/to any affected individuals such as patients or other health related professionals.
NOT be used as a weapon or a means to facilitate direct injury to people.
NOT violate internationally accepted norms when gathering or using information for surveillance/monitoring
NOT violate widely accepted principles of international law and human rights.