Interview with Dr. Michael Johnson

Kimia Vahdat North Carolina State University

Dr. Michael Johnson Dr. Michael Johnson is the professor and chair of the Department of Public Policy and Public Affairs at University of Massachusetts in Boston. His main research interests lie in decision support systems to improve nonprofit organizations and government agencies’ operations and strategy design. We have invited Dr. Johnson for an interview to shed some light on ways in which the OR and Analytics community can contribute to research in the racism and discrimination space.

What is an equitable decision model or decision system in real life applications?

I want to introduce the word “effectiveness” as well as “equity”. Many decision models have industrial motivations. So, for many of the most elementary examples of our applications, the primary goal of the model can be expressed in a straightforward way and quantified using measures that are easily available in the world around us. Also, there’s usually not much confusion about why we wish to minimize miles traveled or maximize throughput. Surely, industrial engineers have thought very carefully about the precise way to do it. For many of these operations research examples, humans I would say are not at the center of the systems, which we wish to build or optimize; of course, trucks require drivers and machines require operators, but in the traditional conception of these problems, the truck drivers and machine operators are not considered to have real agency. For many social policy and urban affairs areas, humans are at the center of experiencing the problem, and in principle taking an important role in responding to it. In addition, understanding these problems requires a conception of social impacts of outcomes that do not necessarily translate well to straightforward quantities that are associated with efficiency measures.

Housing is a great example; the question is: where and what kind (e.g., rental or owner occupied) might we wish to propose new housing to be built or rehabilitated? There are many dimensions of outcomes associated with housing for individuals and communities that don’t align particularly well with traditional efficiency measures. When there is a public policy or social policy dimension to our work, we need to ask, what the impacts are and how we could associate decisions that could be prescribed through a decision model with the social outcomes that we want? So, measuring them is hard and that’s where the effectiveness comes in. How do we know that locating housing of a certain type of this many units in this neighborhood as opposed to another neighborhood will have a better social outcome? There’s also the matter of equity or fairness. When we talk about housing for low- and moderate-income people, that housing often must be subsidized in various ways to make it affordable to these individuals in the U.S., we have entrenched structures of inequality, bias, and racism, that result in many people believing that so-called affordable or subsidized housing, by its very nature, is bad for their neighborhoods. So, from the perspective of low- and moderate-income people, they would like housing that is in the most opportunity-rich neighborhoods that enable them to have the best opportunities for a productive and enjoyable life. So that’s what fairness means to them. But for many middle-class Americans, who believe that the best neighborhoods are the ones populated by people like them, and who perceive new residents who have very different social backgrounds from them as a threat, the introduction of affordable housing would seem to them to be manifestly unfair and should be avoided. Therefore, for this kind of model, we have competing notions of equity, which makes it very hard to agree on a single objective to meet.

In a very simplified example, can you tell us about the work that has been done in this area to give us more insight?

This example is inspired by research I am doing where I’m having to read historically important papers that have had public policy impact, authored by Ed Kaplan, of Yale School of Management. One such paper came out in the late 90s, in the journal operations research, sought to understand how new policies on the lawful exchange of injection needles might reduce the incidence of highly communicable and deadly diseases, such as HIV AIDS by drug users. Kaplan’s insight was that a particular sort of modeling framework focusing not on what people said they did with the needles, because self-reports can be unreliable, but relying instead on documented history of the needles themselves such as, when they were distributed, how long they were out, when they were returned, etc., can help us understand the trajectory of needle usage and to estimate whether needle exchange policies might actually reduce the risk of needle sharing and the exposure to the HIV AIDS virus. So, this is a kind of research that one could think of being focused narrowly on a technical question: how to properly model the life cycle of needles? It’s finding support for a policy that even now is quite contested and controversial. And focusing, not so much on efficiency measures, but on public health outcomes, in other words, effectiveness, and taking the needs of the most vulnerable populations in our country, drug users at high risk of sexually and otherwise transmitted deadly diseases, seriously, and putting their needs and concerns at the center of the model. This is something that means a lot to me personally, and I hope that there can be more exciting research in the future.

What do you think is the responsibility of the OR community regarding DEI initiatives?

In the case of INFORMS, I like to distinguish between the discipline of OR/analytics and the profession of OR/analytics. I think of the discipline as the collective body of knowledge of OR/analytics and related fields that researchers contribute to, teachers draw from in their instruction, and practitioners use in their daily work. However, there is also the profession of OR/analytics, which to me represents the jobs that are done, using various operations research and analytics tools and methods. So, the profession might encompass researchers, professors, analysts, and managers, and also encompass agencies such as, government organizations, and professional societies, like INFORMS. Oftentimes, we think of DEI within the context of the profession. First, we want our organizations to be more diverse and to be more supportive of the careers of historically underrepresented or marginalized populations, such as women and people of color. We want to enlarge the pipeline of future professionals who can come into our organizations and our schools. We also want to make our organizations more welcoming and supportive of them so that they persist and excel. Additionally, we want their professional successes to perpetuate so that organizations can increasingly better reflect the communities they serve and make a greater impact on social justice. That’s what I believe DEI can do for the organizations we work for and for the society in which they’re a part.

The DEI for the discipline is about answering questions such as, “what is there to know about research in the OR/analytics tradition that can help us generate more salient insights about policy interventions and organization designs in which issues of diversity and inclusion can play a central role?” There is a significant amount of work in this area. These studies can provide an evidentiary base for various kinds of policies and interventions. In other words, taking as given that on purely justice terms alone, interventions that can increase diversity and inclusion are worth doing, there’s also the question of are they efficiency-enhancing? That’s an unsettled question. Our society does not exist and does not serve its people without a firm commitment to justice and the rule of law. However, there’s so much more to our society than justice and the rule of law, and we want our institutions to work to balance efficiency, effectiveness, and equity.

Another aspect of DEI in the discipline is answering what kinds of interventions we could design, if we put the needs of traditionally underrepresented or marginalized people at the center of what we do. So, take law enforcement as an example, in which models generally try to give guidance to law enforcement personnel on where officers should patrol or how cars should be dispatched. The historical data can tell us where reports of crimes have been submitted, but that is not the same thing as where the crimes occurred. They can also tell us where arrests have been made, which is also not the same as the real crime that occurred. This is not the same as identifying those who may have committed the crimes. In addition, we know that historic inequities mean that certain populations, particularly African-American and Latino populations are more visible to law enforcement than other populations, and certain behaviors that are perfectly well accepted in white communities are not accepted in communities of color, surveilled and controlled by law enforcement. So, models focused on prescribing the best ways for police to respond to or intervene in so-called hotspots or to design the most efficient interventions like stopping and frisking may have, in fact, been unattended by severely disadvantaged and historically marginalized populations. Does this mean the models are bad? I don’t believe that they are inherently bad, but I do think that where models are designed without the active participation of communities who are implicated by the effects of them, then there is increased risk that the models could be misspecified, poorly implemented, or their results may be misunderstood. Such models could have the impact of increasing marginalization, and disadvantage among already marginalized and disadvantaged communities. We could remedy this by enlarging our notion of modeling to include stakeholders, and representing communities that are implicated in these models and are affected by these models.

What is your advice for students who want to be more actively involved in the DEI research? How can they start the conversation and, where should they begin?
That is a really good question. I would recommend that students start not with the textbooks, the algorithms, and the math, but instead, look around their neighborhoods, communities, and cities and ask themselves, what about the community or the public policies could be done better. I’ll give an example that I think is familiar to many of us. In many cities, we see people who are ill housed. They may be homeless, or may not have a secure place to live and you might ask what can be done? It occurs to me that our engineering training might be useful here; just as some of us are trained to look at a product and say, this needs to do something different and work better, and the engineers say, how can we build it? Well, maybe there’s something analogous in the domain of social policy related to that, where we might see something in our community that we feel is not working the way it ought to, even if we don’t know exactly why. We could challenge ourselves to say, well, is there a solution? Is there a response that is rooted in evidence that could provide clear guidance to practitioners and decision makers in the form of prescriptions? Do we have some notion of social objectives we wish to optimize, which can get you thinking on what kinds of model-based interventions might be possible? It has to start from us looking around and ask ourselves what could work better, and then do the reading. There may be a model based response that has the potential to improve social outcomes. In addition, there might be an opportunity for us to be part of that solution.

What do you think are the next steps for the OR/MS community to develop towards a more equitable organization?

I think we need to be honest and frank about some very basic issues regarding our practice. There was a radio show that interviewed a writer for The Atlantic recently, who wrote a book on inequality in higher education. He was explaining how he got his college education in Alabama at a historically black college, but he noticed not far away from their campus is a campus of the University system that was much better endowed with bigger libraries, better quality housing, greener lawns, etc. The writer reflected that, in many universities like these, even in states where there is a significant fraction of the population that is black and brown, the population characteristic is not reflected in the classrooms. Therefore, I would encourage those of us who are in the classrooms learning or teaching, to ask ourselves, who is in the classroom and who is not in the classroom? If your classroom has 1/3 women, but you know that the population of the state in which your university is located is half female, it’s an opportunity to ask yourself, why is there this underrepresentation, what can we do about it, and what could that mean for the kind of knowledge that gets taught the kind of knowledge that it’s generated through research? So, I really want students and younger people to look around and identify where there might be instances of underrepresentation, inequality, or injustice in various ways. Ask yourself, how can we fix it? What is the nature of this inequity? Where did it come from? And how could we think of remedying it? Then maybe, is there a role for doing so with tools that I’m being trained to use in OR.