A risk assessment response to #ShutDownSTEM
On June 9, 2020, I became aware of the trending hashtag #ShutDownSTEM and a plan in many quarters of the academic and technology world to take a global stand against anti-black racism in science. This was to take place the following day on June 10, 2020. While it was clear to me that the intent was to directly participate in what is now a worldwide protest against institutional racism, police brutality, and other serious social issues that have been bedeviling society for centuries, I felt that rather than addressing these issues by spending a day away from STEM-related activities, it would be far more constructive to instead spend the day fully embracing STEM-related activities that have relevance to the ongoing debates around societal racism occurring in the United States and around the world.
It didn’t take me long to decide to do not one thing but two things on #ShutDownSTEM day. First, share with the public my findings in a project I completed four years ago, but that has relevance today given the current public interest in reforming the way the police in the United States ensure the public’s safety. Second, to show how a basic concept from my area of expertise, aviation risk assessment, can be used to illustrate an emerging concept of racism as a public health issue.
Racial disparities in Boston’s “Stop and Frisk” program
The Boston Police Department uses proactive policing tactics, including the Field Interrogation, Observation, Frisk, and/or Search (FIO) program to disrupt criminal activities. This program was similar to the more well known “Stop and Frisk” program in New York.
In my detailed analysis of that data, which covered data from 2011–1015, two things stood out. The first was an inability to answer a basic question, whether citizens of a particular racial or ethnic group were more likely to be stopped and questioned. All of the police officers had a unique identifier, so I was able to determine how many reports each officer submitted. However, the data provided by the city of Boston did not provide any sort of unique identifier for those who were stopped. There were 143,073 FIO reports that identified the subject of the reports by race, of whom over 61% were African American, but without the ability to uniquely identify a subject, it was impossible to say if any identifiable racial or ethnic group was being disproportionately stopped and questioned.
The second thing that stood out was a bit of a surprise. Two groups of officers had a very big difference in the racial profile of who was questioned. The 20% of the officers who submitted more than 80% of the reports were slightly more likely to question African Americans. The other 80% of the officers, who submitted less than 20% of the reports, were less likely to stop and question African Americans and much more likely to stop and question people from other racial groups, particularly Asians.
Updated Boston FIO records available for analysis
Anyone who would like to do their own research on this data should visit the Boston Police Department page where this data can be downloaded and analyzed. Data from the years 2011 to 2019 is available and can be found at https://data.boston.gov/dataset/boston-police-department-fio
Using an aviation risk model to analyze the effects of racism
In a June 4, 2020 opinion piece in the Washington Post, which addressed several issues related to the death of Geoge Floyd at the hands of the Minneapolis, MN police less than two weeks earlier, Dr. Michelle A. Williams, and epidemiologist and dean of the faculty at Harvard T.H. Chan School of Public Health, and Jeffrey Sánchez, a former Massachusetts state representative and lecturer at Harvard Chan, noted that some commentators held that underlying health conditions may have contributed to the death of George Floyd at the hands of the Minneapolis police.
Williams and Sánchez, noted that the death had already been ruled a homicide, and stated that any explanation that included underlying health conditions was an attempt to deflect blame away from the police officers who ended George Floyd’s life.
They also noted that black Americans suffer from higher rates of diabetes, hypertension, asthma, and heart disease than white Americans, and that they are also more likely to be obese and get to other conditions that can to such health issues.
In a key passage in that opinion piece, Williams and Sánchez state that:
A growing body of literature shows that social determinants — otherwise known as the conditions in which we’re born and in which we live, work and play — are key drivers of health inequities. For generations, communities of color have faced vast disparities in job opportunities, income, and inherited family wealth. They are less likely to have housing security and access to quality schools, healthy food, and green spaces. All these factors undoubtedly undermine mental and physical well-being.
Reading that paragraph made me think of the “Swiss cheese” model of accident causation, made famous by British psychologist James Reason, which is widely used in risk analysis and risk management in aviation, healthcare, manufacturing, and other fields. In short, it models complex systems such as an industrial process as one where bad outcomes like accidents are prevented by a series of protective systems or defensive layers that eliminate the possibility of a negative outcome, make that outcome less likely to happen, or reduces the magnitude of a negative outcome.
As the visual depiction of the Swiss cheese model illustrates, no defensive layer is 100% effective, and for a bad outcome to occur, every one of the defensive layers will have to fail in some way.
In aviation, it is rare for a single hazard to result in a negative outcome. Typically, one or more hazards have to combine in some way to result in an outcome that includes damage, injury, or death. .These hazards can be completely independent and occur randomly, or they may happen in combination.
Understanding risk in the context of George Floyd’s death
One way to understand and model risk is as an equation with three components. On one side of the equation is an unwanted outcome. On the other side is a combination of two things, a hazard and the likelihood of that hazard occurring. In short:
Unwanted outcome = (Hazard)(Likelihood that Hazard)
This model applied to the case of George Floyd, the three parts of this risk model may look like this:
- Unwanted outcome — Death at the hands of law enforcement
- Hazard — Use of a police procedure that may result in the death of someone in police control or custody
- Likelihood — The probability that a particular person is exposed to the hazard
If both the hazard and the likelihood or probability of that risk can be objectively defined and measured, then two different people, starting with the same definitions and values for both the hazard and the likelihood or probability of that hazard, should come up with the same result.
In the case of the death of George Floyd, the hazard was the use of a restraining procedure, including kneeling on Floyd’s neck and back for eight minutes and 46 seconds, that interfered with his breathing and apparently led to his death.
In my opinion, the third part of this risk model, the likelihood or probability of that restraining procedure being used, is at the core of the protests that have erupted not just in Minneapolis, but also in cities and towns across the United States and around the world. The issue is whether being African-Americans carries with it a higher likelihood of experiencing this particular procedure, or for that matter any other police procedure or action, that has the potential to result in death.
Using a risk management approach to understand racism
Williams and Sánchez made observations that when combined with Reason’s Swiss cheese model can be used to characterize social and economic conditions that could result in negative outcomes often associated with racism. Using this model is consistent with the general guidelines for a risk management process followed by US government organizations like the National Institute of Standards and Technology.
Williams and Sánchez pointed out that the following hazards are both associated with differences in health outcomes and more likely to be associated with African-Americans:
- Heart disease
- Air pollution
- Race-based discrimination
- Insufficient sleep
- Law enforcement encounters
Each of the above conditions could be thought of as the hazards (a red arrow) in the Swiss cheese model. The following, either explicitly mentioned or implied by Williams and Sánchez, can be thought of as the protective mechanisms (the Swiss cheese slices) in the same model:
- Housing security
- Fair job opportunities
- Inherited wealth
- Quality schools
- Healthy food choices
- Access to green spaces
Following the logic implied by Reason’s Swiss cheese model, if someone wanted to reduce a particular negative health outcome, for example, death at the hands of law enforcement, there are two key questions to ask:
- What characteristics are associated with being exposed to one or more of the hazard conditions?
- What characteristics are associated with benefiting from one or more of the protective mechanisms?
Risk models as a starting point
It has been my experience in aviation that building a risk model provides those involved with addressing that risk a guide to systematically build an objective understanding of the risk. While such a process does not eliminate the need to address the subjective or emotional aspects of a risk, it does allow those who are dealing with that risk a way to identify both opportunities to reduce or eliminate hazards and opportunities to create protective mechanisms that can keep unwanted outcomes from occurring.
In has also been my experience that building a risk model is a small part of the risk management process, serving primarily to create a detailed understanding of the elements that contribute to a top-level risk, an understanding that often provides a guide to creating a process to systematically reduce or even eliminate the risk. If one starts with a societal level risk issue, whether it be racial disparities in police behavior or air transportation accidents, change may take years or even generations, and in both areas understanding the problem in a systematic way is a good place to start.
Field Interrogation and Observation Program — data.boston.gov
Guide for Conducting Risk Assessments — National Institute of Standards and Technology (NIST); Special publication 800–30, Revision 1; September 2012
Observed racial disparities in the Boston Police Department FIO program — Todd Curtis, 8 February 2016
Racism is killing black people. It’s sickening them too — Michelle A. Williams and Jeffrey Sanchez, Washington Post, 4 June 2020
Revisiting the “Swiss cheese” model of accidents — EUROCONTROL Experimental Centre, 2006