The Ethics of Visualizing during a Pandemic

I’ve had a number of conversations this month about ethics in pandemics. Here’s probably a good starting point: people are scared beyond measure and don’t know what to expect.

Uncertainty fosters speculation, calculation, and various emotional responses. We are all worried how this will affect us and others directly and indirectly. At times, we have competing interests or concerns. Yet, one surprising upside is disasters generally foster greater acts of solidarity. Many of us want to support others and our skill resides in the realm of data. Here’s how to help without accidentally hindering.

A Starting Point

In a pandemic, we have a lot of insecurities. Fear is the most common emotion driving the conversation, and to a smaller extent anger, resentment, and fatigue (usually from those more directly affected in some way). Robert Sapolsky, a primatologist and neurobiologist, has found stress hormones skyrocket when we live at the bottom of a social structure, but also when our social position is threatened. Pandemics threaten us bodily, economically, socially, and politically. So often, when we analysts look at the data, we’re not factoring in all of the above. Rather, we’re extremely focused much closer to home: on how to quell our own fears and questions. As such, we’re less likely to pause and consider our actions, making us far more likely to accidentally and unknowingly create harm in the process. We’re not rational when we’re afraid. We’re too focused on survival and reacting.

Ethics require pausing. They, too, come with a ton of uncertainty – not something many of us voluntarily want to add to an already full plate of ambiguity. Ethics also force practice and dialogue, which often means picking a less desirable answer than the one we want. Further, they require admitting that what we want (answers! certainty!) may not be possible and can hurt others in the process. Ethics are philosophy, also less exciting in an age where we trust in data, not what feels like abstruse riddles.

While data exists in shades of grey and nuance, it’s easier to focus on the Boolean aspects – I do X next or the data definitively says Y. We analysts may caveat the assumptions, but our end users might not. Data literacy is also an ethics issue.

A key part of data literacy is defining. The more we do this for our ourselves and others, the more likely the information itself is usable and clear. But, defining isn’t a list of disclaimers at the bottom: it’s a process that interrogates the data and our intentions. This process looks at motivation and ensures we aren’t potentially obfuscating valuable information for our gain. One key metric in pandemics is cases, yet how we define a case is widely varied and often blurred. Even deaths aren’t as clear cut as they seem: they often only count deaths directly associated with the pandemic that are diagnosed. They may or may not count deaths at home, depending on reporting entities, testing, and how the death is logged. Peripheral deaths, those from hospital overload, aren’t counted. Nor does it take in account the personal toll from these deaths – the lasting effects on families, friends, communities, and the staff providing care.

Data literacy within an ethical approach requires us to take ownership of mistakes in understanding. It’s far too easy to write off someone’s misaligned conclusions of our work as their problem. It’s not: we as creators own that. It is our job to put in enough guardrails to keep others from falling off a cliff. Other professions with ethics, such as the medical profession, clearly work to ensure patients understand risks of treatment and non-treatment.

But, data literacy isn’t the biggest complication. We are.

A Look at Morals

A crucial part of where we’re struggling is around morals. Morals are horribly personal and we often don’t realize they exist until something else forces us to question them (such as this example). We act on them without thinking and many societies make it easy for these to go unchecked, particularly for persons who are privileged in various ways. When we don’t check our morals, it’s hard to have productive conversations in this space. We’re experiencing this now in our discussions on how and if we visualize this topic, what data to use, and what domain expertise is needed.

Some values at play involve sanctity of life, autonomy (disguised in part as rights), justice (disguised as rights as well), beneficence (harm reduction), and confidentiality (trust). The sheer number of conflicts makes it hard to see what’s happening. How do we prioritize each of these in decisions and what nuances matter? We struggle with these conversations even in the best of times. Times of stress add to the struggle and uncertainty. Few are trained on naming these values, yet identification helps find the conflicts.

Pandemics require strong acts of solidarity through specific, time-sensitive changes in behavior for the benefit of the public. This often means limiting autonomy (such as by stay in place orders) to support sanctity of life (fewer people dying or experiencing long-term complications), beneficence (less harm to those most vulnerable), and preserving trust in information (confidentiality). Justice factors heavily into these discussions as to how we make this possible and what populations are and aren’t protected. We’re seeing these exercised in how countries handle the pandemic across all fronts.

Philosophers have long discussed how right and wrong plays out in societies. What happens when our values conflict? Thomas Hobbes, for example, has argued we can be self-serving in our interests. Philosophers like T. M. Scanlon push us to think beyond ourselves. Individual morals will have a significant impact during this pandemic. We saw this play out in the influenza pandemic in 1918 and we’re seeing these discussions again in modern times.

Beyond personal morals, professional ethics also come into play.

An Interdisciplinary Discussion

A number of people are trying to solve what COVID-19 means to them personally. Unfortunately, the answers are far from straightforward and visualizing this data ethically requires knowledge across a number of domains. We’re running headfirst into conversations occurring in:

• virology (modes of transmission, life on surfaces, physical vulnerabilities, mutations, silent period)
• epidemiology (what social conditions affect spread)
• public health (how do we best service those seeking healthcare, how to ration, how to prioritize)
• medicine (how to treat, how to keep providers safe)
• politics (how information is shared, what’s tracked publicly and what’s not, prioritization of funds, denizen management and policy, allocation of economic supports, borders, trade)
• demography (what populations are most effected vs protected)
• economics (highly sensitive and reactive to the above)

This is by far not a complete list, nor is it any indicator of order. Rather, this list showcases the broad amount of expertise needed to even begin to understand the scope of what a pandemic means, pointing us to areas of research to begin considering this information ethically. Ethics create a ton of work intentionally to slow down our thinking and consider broader angles and implications. Each of these domains often bring their own ethical values forward: priorities can and will clash cross-discipline.

When we look at how the virus spreads, transmission factors into discussions of how to reduce exposure for the general population, but also frontline staff and supports treating those already ill. Public health helps advise where we are on the curve and how we can best mitigate with several domains trying to protect and bring attention to those most vulnerable. Policy often dictates how these measures are implemented, but also supported.

Pandemics expose inequality in stark ways. Employees who are able to work remotely face different moral choices than those who cannot perform their duties from home. Precarious populations that live paycheck to paycheck face very different moral considerations than those with enough money to last several months. A subset of workers are also part of how this pandemic is mitigated: medical providers, but also those cleaning hospitals or providing delivery of goods, in addition to a number of other services needed for continuity. Almost everyone has something to lose, but what is lost is far from equal.

Those with the smallest risks are often more willing to take risks in other areas. Removal fosters indifference: it’s only X number of deaths comes from a position of removal. Just this population also comes from distance. Many in the disabled community fear for their lives, not directly from the illness, but from a potential deprioritization in the value of their lives during a crisis.

When morals ram headfirst into insecurities, you see them get compromised. If your job is on the line and you morally disagree with something, you might fuzzy up the line to make something “okay” because you need to get by: it’s common to adjust a moral line when values conflict (security over harm to others, for example). Ethics require bravery and bravery typically means sacrificing. We celebrate bravery because it’s rare.

Visualizing Ethically

Some of the best visualizations that are ethical come from those most affected. Consider this visualization made by Kate Brown, a cancer survivor:

Most iterations on this data set dealt with rates and made various charts over time. The most obvious part of practiced ethics is the output: Kate addresses head-on what a patient is likely to experience. She has empathy and tells her end user something vital: you are not statistic. But Kate also does something key to guiding users ethically through the whole cycle of data literacy – she tells us explicitly that this data does not represent all the data for this diagnosis. She provides nuance and a place to seek more information: a doctor and personal treatment plan. This is a fantastic example of ethical guidance.

Pandemics require this type of framing and commitment as well. By the end of this pandemic, we will all likely know someone directly affected by COVID-19, if we don’t already. How we choose to communicate about it matters both now and later.

Here are some tactical questions to ask so you can check ethical framing.

Data Questions

• Are you disclosing the limits of your data in a way that is clear? For example, what all influences how cases are reported?
• If you’re visualizing information that is time-sensitive such as cases, are you willing to commit to daily updates and for how long?
• Are you disclosing all sources?
• How does storage affect updating and relationships?
• What inaccuracies exist?
• What transformations are you making in storage?
• What does NULL mean?
• How are you bringing sources together?
• What gets left off in the case of joins?
• What decisions have you made in storage? (ex: how locations are classified?)
• Does your data involved mixed grains (ex: country vs city level data or a blend of both)? (Side note: medical geographers might have something to say about this.)
• What data aren’t you showing that’s a factor? (ex: stay in place orders)

Visualization Questions

• How does the visualization limit what you can understand? (ex: A line chart shows trend, but not potential geographic patterns)
• What expertise is required to follow it? (ex: log scales)
• Who have you considered as your audience?
• Who have you (intentionally or unintentionally) ignored?
• What is your motivation for doing this?
• What values are you espousing in your visualization? Do they support or conflict with other values?
• How can your visualization be used for harm? (If you must, find someone who loves to play devil’s advocate)
• What would it take for you not to publish or share this visualization? (If the answer is “nothing” you need to check that)
• How can someone use this to: 1, defy pandemic policies, such as stay in place orders 2, promote or perpetuate racism 3, diminish lives of those that are disabled 4, conclude falsehoods?
• Does your design diminish the topic? Consider how it’s interpreted without reading the words.
• Do you create overt alarm? Most people are already stressed.
• Does it overly abstract and calm too much? Some people aren’t taking this seriously enough.

Publishing Questions

• How many visualizations already exist on this topic?
• What are you adding to the conversation?
• What harm happens by waiting to publish?
• Why must you publish this now?
• What is your motivation? Attention is a form of currency, so consider that carefully.
• What actions will someone take from this?
• Are you explicitly making calls to action accessible?
• How will this harm others?
• For someone harmed by what you make, how will you justify or explain it?
• Have you read 10 Considerations before You Make Another Chart about COVID-19?
• Have you considered the rights of others?
• Are you willing to take responsibility for misinterpretations? (Consider what expectations of journalists are.)
• Are you willing to take this down or issue an apology if the data is false, conclusions misrepresented, or otherwise misinterpreted? (Scientists issue retractions when need be.)

Value Assessment

Sadly, even all these question don’t cover everything. There’s still a ton missing. This is part of why ethics feel challenging – it’s hard to be purely tactical with them. It’s normal for ethics to feel ambiguous and they should induce a level of doubt into a decision.

Some tools help identify values. They are best found in codes of ethics and defined within the document itself. When looking at challenging ethical situations, try to name the values. If you’re struggling, there’s likely multiple. Then see where you fall and perhaps try this with others as a framing exercise. Ask questions that force nuance: to what level or extent? Are there caveats or limits to your answer or is it always so? (Hint: there’s almost always caveats.)

For example, accuracy represents trust, justice/fairness, and autonomy. If you have false information, it risks violating trust, preventing a fair decision, and removes the autonomy of making an educated decision, By removing facts, you force a decision on others. It’s why we analysts care so much about accuracy.

Make a few copies of this tool and see where values shift and what nuances force them to move depending on the question. Is there a time you’re comfortable being less accurate, for example? Try to understand why. By exploring, you’ll begin to understand your fundamental morals, which will influence how you practice ethical decision-making. You can even make awesome vizzes on values.