Best practices for using data during a crisis

Palantir
Palantir
Mar 17 · 7 min read

The COVID-19 pandemic appears likely to rank among the most serious epidemics in modern history, with a mortality rate and global impact that has invited comparisons to the Spanish influenza of 1918. There is reason to hope, though: Data, technology, and globalization have fundamentally changed our capacity to combat such a threat.

But we would do well to remember that, in times of crisis, societies are judged by how they rise up to protect their people without sacrificing their values. Our institutions must be thoughtful and vigilant in their use of technology in the coming weeks and months. The precedents we establish today may determine how humanity responds to global threats for decades to come. It’s critical that our response is proportional to the threat, supported by effective technology and comprehensive data, and most importantly, guided by the values we hold dear.

For 15 years, Palantir has helped institutions use data effectively yet responsibly. Below, we recommend a set of principles that organizations combating COVID-19 should follow to maintain this balance while using data science and technology to advantage in their response.

First things first: Data isn’t a panacea. As appealing as it may be to believe that data science alone can solve this, or any other crisis, complexity and global future implications require human judgment and expert knowledge. Data is simply a tool to empower wise, and politically accountable, judgment and knowledge generation. Like all tools, data and technology have appropriate uses and limitations. Knowing how to competently apply data science to the right set of problems will serve as a critical asset for augmenting and enhancing comprehensive strategies to battle this public health crisis. The following concepts are meant to help inform and guide the use of data-driven approaches to enable a successful response to the current pandemic.

  • Focus on decisions to be made, not just insights to be discovered. The only data that matters in a crisis is that which can be leveraged towards targeted, actionable outcomes or provide necessary context to make consequential decisions. This is the wheat. In a data-rich environment there are an endless number of potential avenues of analysis, and time will be wasted if all are pursued without a focused strategy. These interesting — but not necessary — inquiries are the chaff that should be put aside. First, identify crucial decisions that must be informed by data and focus on analyzing only the data necessary to enable these decisions. Vital COVID-19 decisions might include: How do we maximize hospital capacity and function? Which containment strategies do we pursue, and where? How do we preserve the supply-chains of essential goods?
  • Start with the data you have. The natural temptation is to amass all data that seems to have potential value, on the hunch that better decisions will emerge from a more massive stock of data. Instead, begin by assessing what data already exists and is readily available for use. This allows institutions to understand not just the wealth of their existing data assets, but the utility, structural features, and quality of that data. From this, ask these questions: Where are the gaps? Who has what data? What’s working? What’s not working? What are the known limitations? Only after identifying critical gaps should an organization reach for more data — particularly where doing so may harm individuals’ privacy or civil liberties.
  • Invest in data management; beware the shiny new object. Robust data-management infrastructure provides the critical foundation for more advanced applications of data. This centers around an ability to evaluate data quality, to curate it with critical contextualizing details, and to ensure its recency. The foundation for informed, data-driven decisions involves trusting data, preventing misuse, and protecting the individuals whom it concerns. Strong data-management may prove to be a markedly more valuable and actionable investment for building out critical analysis than rushing to adopt speculative and poorly understood novel technologies and applications. For instance: social media sentiment analysis applying advanced AI techniques (e.g., Natural Language Processing) may be a tempting method of measuring population response. But absent a specific strategy about what will be done with that analysis, why it’s valuable, how accurate or reliable it might be, whether it contains bias, etc., this pursuit may serve more as a distraction (or worse) than as a meaningful solution to critical challenges in a crisis. When technological gaps are identified and new approaches are considered, remember that technology is a means to an end: Start with the problem to be solved, focus on efficacy versus downside threats or problems, and work from there to the simplest technical solution. It is always tempting to answer crises with the newest, and often most intrusive, tech. Resist that temptation and concentrate on the demands of the problem at hand.
  • Look beyond quick wins: Have a data strategy. Think broadly and plan ahead to build a unified data asset that can be quickly and proportionally adapted to address new concerns, many of which may be interrelated. A challenge on the scale of a global pandemic demands a multifaceted and holistic response and immediate objectives of minimizing the spread of the virus must be pursued in concert with plans to maximize hospital readiness and ensure the integrity of supply chains. If structured thoughtfully, for example, investments in national supply-chain analysis may be quickly repurposed to anticipate hospital shortages. A data strategy formulated to articulate and optimize appropriate secondary uses will enable multiple positive outcomes and ensure that necessary data protections (e.g., consent, minimization, aggregation, de-identification) are in place for onward uses of sensitive data.
  • Set the rules of engagement from beginning to end. Rich data sources often inspire unanticipated — even rogue — analyses. Establish and enforce collective ground rules on how the data should be used and who should have what levels of access to, and use of, data. Misuse of data can result in public mistrust in institutions. Even the most well-meaning of problem solvers sometimes are blinded to the risks of the solutions they create.
  • Establish safeguards to maximize correct decision-making and human accountability. Prior to taking any significant actions based on data, technological and procedural protections should be embedded in new systems and decision-making processes to protect the validity of input data, the quality of analyses, models, and logic applied to that data, and the integrity of decisions based on it. Individual human beings must be accountable for all decisions potentially affecting individuals’ health, safety, privacy and civil liberties. This is even more important in situations — like the current pandemic — in which the stakes are high and time pressure is a significant factor.
  • Secure your data before sharing it. In a crisis, public institutions often must share significant volumes of highly sensitive data under extreme time pressure. The sharing and consolidation of information, however, greatly heightens the risk to such data (such as offensive cyber operations against public infrastructure and accidental health data exposure. Already, scammers have, among other exploits, counterfeited a widely-used COVID-19 mapping site to launch attacks against those simply seeking information). Appropriate and proportional data sharing often is decisive in a successful, cross-jurisdictional response. But once data is distributed, it is difficult to take back and lock down. To the greatest extent possible, technology should enable data sharing only to the extent needed and, critically, only for the time period needed. Start with a clear assessment of data sharing benefits and risks and then define mitigation strategies to enforce security, privacy and data minimization measures while enabling the success of the mission.
  • Build a data governance body. Accountability and trust around secure and appropriate use of sensitive data is never more necessary than in a crisis. To balance accountability with the speed decision-makers require, a dedicated body of experts should be used to advise on managing data access, reviewing applications, and recommending measures to mitigate the potential for misuse. Such a body should not constrain action but should temper it with broad subject-matter expertise and a sensitivity to protecting individual liberties. It should include mission action officers, technical specialists, and experts in privacy, civil liberties, ethics and compliance to ensure that all the relevant equities are considered.
  • Serve the patient and respect their human dignity. Societies must emerge from a health crisis with their values intact. In some cases, totalitarian responses to pandemics may be more “effective” in the short term. Draconian measures have often been exercised before by societies at the expense of individual privacy and liberty. In the 21st century, this does not have to happen. Responsible pandemic responses should begin with a commitment to preserving core societal values as part of an optimized approach and not in conflict with it. Especially at a moment of heightened volatility and political polarization, public health approaches undermining individual rights are not only unjust but risk exacerbating social tensions already at risk of boiling over into antisocial behavior.

Finally, we all must be mindful: History has its eye on us. Exceptional actions carried out under exigent circumstances often define new norms, for good or ill. Permissible exceptions to longstanding norms should be explicitly established with reference to the attendant risks and tradeoffs, but also with a fixed point in the future for sun-setting or, at minimum, re-examining. We must not blindly accept the mantra of “desperate times call for desperate measures,” but instead forge solutions that can survive a return to normalcy and not fundamentally alter our societal values. Any exceptional measures must be clearly justified by the facts and conditions of the moment but, also, in enacting them, build in mechanisms for rolling them back after the crisis and soberly evaluating the extent to which they were necessary and how we can do better next time. This is an emergency — perhaps the defining one of our age. In acting decisively to defeat this pandemic, we must do so in a way that we will recognize ourselves when it’s done.

Author

Courtney Bowman, Palantir Privacy and Civil Liberties Engineering Lead

Palantir Blog

Palantir Blog

Palantir

Written by

Palantir

Palantir Blog

Palantir Blog

More From Medium

More from Palantir Blog

More from Palantir Blog

More from Palantir Blog

Code Review Best Practices

13.6K

More from Palantir Blog

More from Palantir Blog

How we do Agile

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade