An AI Ethics Checklist for COVID-19 Healthcare Technology
As the confirmed cases of the COVID-19 virus continue to climb worldwide and many of us go into the end of our first full week living a new lifestyle grounded in social distancing the question is being raised with increasing urgency: is the end to this pandemic in sight?
To solve the puzzle knot of this pandemic there will need to be a coordinated effort across many industries, governments, and scientists. Though it is important to name that implementing any solution will take social, economic, scientific, and other multi-pronged strategies in many different contexts, it is clear that part of the solution is going to come from the technology sector. More specifically, one of the sectors that has been able to make the most progress in addressing the complicated web of issues emergent in the spread of the virus has been the Artificial Intelligence technology sector.
However, much like any discussion about artificial intelligence technology development and implementation, when it comes to addressing the corona virus there are several vital ethics questions that must be asked. If companies, governments and research agencies do not begin to ask some of these ethical questions in their work there may be harrowing concerns further downstream beyond the spread of the virus.
On March 13th Forbes published an article entitled “Coronavirus: How Artificial Intelligence, Data Science And Technology Is Used To Fight The Pandemic” written by Bernard Marr. In this article Marr lists 10 ways in which AI, ML, and DS are being used to address the pandemic. I would like to use Marr’s list as a framework to propose questions (3 questions per item) that folks designing and using this tech to engage with the virus might ask themselves to better ensure ethical creation and application. You can think of it as an AI ethics checklist (a very original idea!). Please note that this is not comprehensive and is simply a list to get the conversation started as we enter this new conversation about how we stay ethical while we develop technology to address a rampaging pandemic.
Using AI to identify, track and forecast outbreaks of the virus
Which reports, platforms, and documents are you drawing from? Are those data sets representative of the global population? If not, is it clearly published why the data set is more limited?
When scraping social media platforms, do you have a clear policy around privacy, consent, and storage of data? How does your privacy policy relate specifically to Patient Health Information (PHI)?
Is your predication algorithm trained off of culturally-relevant data to the area you are predicting in? Are there more effective ways for your coding team to understand the relevant socio-cultural factors that might impact the spread of the disease in the areas you are predicting for?
Using AI to help diagnose the virus
Can your AI technology adapt to different cultural contexts in a way that is accessible? Have you designed your AI technology in a way that speaks across different languages, assumptions, and tech literacy?
Is your diagnosing apparatus or algorithm available to those who need it or is it hidden behind a pay-wall or other form of red tape that make it difficult to access in the field?
What is being done with the primary data being secured through diagnosis? What about the exhaust data? Is there a way to use differential privacy or a similar method to make your data available for the common good in a privacy-conscious way?
Using AI to process healthcare claims
Does your AI technology not only address efficiency of processing claims but justice issues related to access, ease, and information translation?
Is there a way for your business model to assist those who cannot afford health insurance but are in need of immediate care?
Is there a way to work more closely with patients to determine useful adaptations for your technology?
Using AI embedded in drones to deliver medical supplies
Are you being intentional about ethically sourcing your materials for drone building? Including precious metals used for CPU/GPU chip manufacturing?
Are there ways to build coalitions with other similarly-minded organizations using drones to deliver supplies? Would this help serve a greater number of impacted individuals?
For drones being used for surveillance of quarantine mandates, do you have a posted list of ethical principles that ground the purview and legal foundation of their actions?
AI robots sterilize, deliver food and supplies and perform other tasks
Has there been an environmental sustainability analysis on your robot’s workflow and macro business plan?
Is there a clear accountability procedure in case something goes wrong in the work flow?
Beyond accountability, have there been adaptations to better serve immuno-compromised populations?
Using AI to develop new drugs
Where are your findings being published? Are those findings open source?
How are drugs being tested? Does your AI drug testing follow the same ethical standards that more traditional testing has followed?
How have you ensured that drugs created through AI technology assistance are available to those most in need of them?
Using AI to develop advanced fabrics to offer protection
Are the fabrics crafted with a clear ethical standard as it relates to production and labor?
Have the fabrics used in face masks and other COVID-19 related medical garb been constructed in such a way that they are accessible, cheaply produced/available to all, and serving the needs of individuals in many different environmental contexts?
Have you made a clear recommendation about how these advanced fabrics should be properly recycled and/or reused without spreading the virus?
Using AI to identify non-compliance or infected individuals
Does your organization have a value statement about individual agency vs. the agency of the state? Has your organization actively affirmed international Human Rights as defined by the United Nations?
How is your AI technology used to support social cohesion/social health? Does that come at a cost? Is that cost worth it?
How does your AI technology balance between surveillance and privacy? What is done with your surveillance data? Is it given to another entity or disposed of after use?
Using AI to program chatbots to share information
What values are embedded in your chat interface? Is it accessible to folks of all abilities?
Have you done an analysis of biases inherent in your chatbot or on its training data set?
Is your chatbot available to everyone? How can you make your chatbot more accessible in its language, purpose, and design?
Using AI algorithms to drive supercomputers working on a coronavirus vaccine
Who is represented on your design team? Are there a diversity of backgrounds in the group that is determining the objective function of the supercomputer?
Which training data sets are you using? Is it possible to diversify training sets using a variety of contexts? If not, is it possible to use resources to create new training sets that implement greater diversity in data?
How can you more effectively and efficiently share discoveries with other teams to collaborate towards faster solutions?
It is unclear what steps must be taken in order to contain and address the complexity of issues caused by the COVID-19 pandemic. However, what is clear is that whatever the solution to this pandemic will be AI technology will play some role in its creation and application. The question for technologists in the field is how to design and use that technology in a way that if effective, efficient, and ethical. My hope is that this checklist provides a basic framework to begin pieces of that conversation.
Stay Radical!
-Dylan Doyle-Burke
