Researchers employing Artificial Intelligence in their systems have been facing more scrutiny from funding agencies and institutions in order to ensure the AI systems they build are "ethical" and will benefit society as a whole.

Consider this passage from resource [3] listed below; specifically, page 28:

"While some suggest the incorporation of AI in health care practice will give physicians more time to engage in empathetic discussions with their patients, others are concerned that AI could reduce the care setting to a mere technical encounter (Nuffield Council of Bioethics, 2018). AI is primarily trained on quantitative measures such as images, health records and blood tests, which runs the risk of reducing clinical diagnosis to only measurable data and failing to capture the nuances of a patient's values or social, economic, and cultural situation and the verbal and nonverbal cues that inform clinical judgment about a particular patient (Future Advocacy, 2018). Furthermore, if AI is fully integrated into care settings, will patients have the capacity to refuse care from AI? Is a patient's right to health synonymous with a right to access health care delivered by clinicians (Future Advocacy, 2018)? And as computational algorithms become increasingly sophisticated and reliable, what role will or should clinical judgment play? Could the diffusion of AI through clinical processes have the effect of displacing clinical experience as the gold standard of knowledge in health (Char, Shah, Nigam, and Magnus, 2018)?

In response to concerns about preserving the relational core of the clinician-patient interaction, it has been argued that the main use of AI should be to complement clinician judgement, not to replace it, and that the development of AI in health care should be undertaken with the clinician-patient relationship in mind (The Centre for Internet and Society, India, 2016). However, this leaves as yet unanswered how discrepancies between AI-mediated outputs and clinical judgment ought to be resolved. Clinician trust of AI outputs is critical if AI-mediated care is to realize its benefits for patients. Lack of transparency about how AI systems work, particularly how machine learning generates a clinically-relevant output, is a particular concern (AI Now, 2017). In addition, if a medical error occurs as a result of applying an AI-mediated diagnostic or treatment decision, who should hold accountability for the error: the AI system, the developers of the technology, or the clinician delivering the care?"

Instructions: As you think about AI and healthcare, your task is to devise a set of 4 guidelines for researchers to follow when exploring AI. To accomplish this, first review the recommendations and best practices as outlined in Resources #1-4 (below). Identify if there is a consensus about using AI in healthcare research. Summarize your findings by preparing a set of 4 guidelines that researchers can use when designing their research involving AI and Health. There is no specific word-count requirement/limit for this Assignment. Use your best judgment to ensure that you provide clear, comprehensive, and well-supported guidelines. All resources must be cited and referenced using IEEE style.

Create the set of guidelines using the following structure:

• Guideline # : title for the guideline.
• Scenario: Describe how the AI is being (or could be) applied.
• Recommendation: Describe your recommendation based on the scenario; i.e., what the researcher should or should not do.

Resources:

(1) IBM Cloud Education (March 2021): AI Ethics | IBM: https://www.ibm.com/cloud/learn/ai-ethics

(2) IBM Research Insights (April 2020), "Advancing AI ethics beyond compliance: From principles to practice": https://www.ibm.com/downloads/cas/J2LAYLOZ

(3) Joint Centre for Bioethics, University of Toronto (July 2018), "Ethics and AI for good health, symposium report": https://jcb.utoronto.ca/wp-content/uploads/2021/03/Ethics-and-AI-for-Good-Health_Symposium-Report.pdf

(4) Canadian Institutes of Health Research, "Ethics at CIHR": https://cihr-irsc.gc.ca/e/29331.html

Be sure to properly reference all citations you use. Good luck!