Dear Data Scientist: A Call for Humility as an Ethical Imperative
Ethical humility for technologists in times of crisis
By Maureen “Mo” Johnson
Mo is a data ethics professional, with a focus on digital cooperation, human rights and public health. You can find her on Twitter at @moridesamoped
In a global crisis, humility is an ethical imperative. Humility allows for listening, collaboration, and the ability to be part of a group effort without receiving individual acclaim and attention. We need many people working together to overcome this global crisis.
—
Dear data scientists and technologists:
If your first instinct is to build a predictive model to fight the pandemic, please pause and read this.
For two years, I led the community for the Global Data Ethics Project (GDEP), a multi-stakeholder data ethics project that aimed to help data scientists consider the impact of their data work. Before I became a data ethics professional, I worked in public health research, as a medical anthropologist and policy advisor, in the field of infectious disease. I offer the perspective of someone who has advised and supported ethical considerations in volunteer data science projects, and it is primarily intended to be read by volunteer data practitioners who wish to support our global community in combating the coronavirus pandemic.
I am writing this because I have seen projects involving predictive models and data analysis that seem to have little context or the support of experts with deep domain knowledge in infectious disease and public health. Though many of the volunteers may have the best intentions, this does not make up for the potential harm these models may cause. Using non-expert models to drive public health policy decisions could be devastating, especially if the implications for the failure or success of a particular model are poorly understood or communicated.
As a techno-optimist but not a techno-solutionist, I feel hope and also advise caution for many technology efforts. Volunteer civic data science and tech projects can offer support on many issues, but not all problems are technology problems, and not all answers are predictive models. Choosing to take on less glamorous tasks, technology-related or not, means that you adopt a team-based mindset and a commitment to offering help based on where it is most needed, and not on what you would prefer to do. This requires humility because you may have to acknowledge that your expertise might not be needed at this time for this problem.
In a global crisis, humility is an ethical imperative. Humility allows for listening, collaboration, and the ability to be part of a group effort without receiving individual acclaim and attention. We need many people working together to overcome this global crisis.
This means that data science can play a role in the scientific effort against coronavirus - it is one part of a collaboration but not the only part. Tools from data science are useful for many initiatives led by infectious disease epidemiologists, many of whom are data scientists, too. We need data experts to provide support for data collection, advise on data security, examine data integrity, and normalize and provide key analysis. However, creating a predictive coronavirus data model in the absence of direction from experienced infectious disease experts is irresponsible. A model is only as good as its assumptions, its data, and the multidisciplinary team contributing to all parts of the data pipeline, including the ethical oversight. Without the support of expert domain knowledge, you may be creating a model just to create a model.
You may read the above paragraph and think you can offer your skills. This is a great intention, but it is better if you have an existing partnership with experts in epidemiology, virology or public health, or if they make a call for your particular skill set. This needs to be said directly: If you are not already connected with an organized initiative or working directly with experts with deep domain knowledge, do not call or write them. They are busy, overwhelmed and working on urgent projects already.
As a project, GDEP is largely retired, though it did spark many initiatives and offshoots. It is my belief that ethical codes cannot be designed to be universal. For example, a singular ethical code cannot encompass some of the adversarial trade-offs between privacy and transparency in every context. However, an ethical code can provide guidelines to help us consider the ethical implications of all data decisions. It can provide questions to consider before building and deploying models and algorithms that affect the lives of many people. For instance, if you have decided to build a predictive model, some questions could include:
Who are you burdening with having to interpret this model?
Will their interpretation lead to policy decisions that will cause harm and kill people?
Do you understand the purpose of an epidemiological model?
Have you imagined and communicated unintended consequences of your work?
Also, if your predictive model may influence future policy and individual decision-making, are you comfortable with the burden that your model may lead to more deaths and harm? If you believe your model only adds benefit, it is time to pause. Uncertainties in your model might be appropriated to fit different political agendas in the absence of context. We live in an information ecosystem where incorrect or badly contextualized information will literally kill.
I understand the drive to help in these unprecedented times. Here is a non-comprehensive list of ways to support the world-wide effort to overcome coronavirus.
Contribute to an existing open-source project that isn’t a coronavirus modeling problem. Also, consider supporting a non-coronavirus project. Many ongoing projects existed before now that need technical support and documentation.
Join existing calls to action. The New Jersey governor announced that the state needs COBOL programmers to support its overwhelmed unemployment system.
Broadcast messages from public health experts who are on the front lines of providing essential and timely information about safety, health and stopping the spread.
I’ll repeat the basics: Stay at home, wash your hands and practice social distancing.
Take care of your emotional health, too. Consider baking -- a recipe is a type of algorithm.
Reach out to someone who is isolating on their own. Ask how they are doing. If you have older relatives who aren’t tech-familiar, you may need to make a phone call.
Direct friends and family towards factual information, and listen to their concerns. Be gentle. We are all scared, and some people have better access to information than others.
Read and share socio-technical works by women, especially black women, to understand the implications of technology. Many of them have been sounding alarms about bias, privacy and power for a long time.
Advocate for more ethical technology at your company. As companies respond to coronavirus and interactions are driven online, emerging and existing privacy issues should be part of the conversation.
If you still have a job, hand out money in tips/donations/anything. Many people are struggling and cash is the most direct way to help them.
These things take work and may seem less glamorous than building a new model, but they are needed and essential. Please be humble and consider where your efforts may have the most impact. We are all in this together.
Good luck and stay healthy.
-Maureen “Mo” Johnson
@moridesamoped