Sean Luo
Hypothetical illustration of a monotonic relationship between resource required to achieve an optimal outcome (Y-axis) and severity of illness/degree of disadvantage at the start (X-axis). Two-phase transitions: In the linear region, a proportional compensation from an external source (say government) is adequate in remediating baseline differences. In the exponential region, linear policy fails and it's impossible for linear agencies such as the government to achieve optimal performance as compensatory actions need to be exponential. In the third region, which in high dimensions can be multiple small regions, the policy is non-monotonic. A small spending in a very specialized way may be required to achieve a better outcome, whereas, increasing spending in a conventional way may worsen outcome. 
Source: Sean Luo

New York City educator Eva Moskowitz created a monster in a figurative sense of the word. One the one hand, her schools produced positive and dramatic results, with black and Hispanic students from low-income families achieving higher scores than wealthy suburban schools. On the other hand, some parents compared her to Mao and Hitler, despite the fact that she is the descendant of a Holocaust survivor.

A recent Atlantic article provides a thorough and impartial review of her work. Although many parents have protested, as this article illustrates, others are voting with their feet to try to get into the schools (10,000 applications for 3,000 spots). 

Why all the controversy? Let me explain.

Moskowitz built a system that focuses on discipline, standardization, drilling, and explicit behavioral modification that works very well for large swaths of underprivileged children who do not always receive this type of training at home. (And while her application of these techniques to inner-city schools is new, the techniques themselves are as old as the educational systems, especially in East and Central Asia.) However, these techniques work poorly for certain children. In fact, some of these children cannot tolerate the techniques and have to be removed from the system altogether.

I can imagine being the parent of a child from the latter group; imagining that I finally found a gateway to ensure my child a quality education, only to find that the system either ejects my child or makes their problematic behavioral patterns even worse. I would file lawsuits.

The fundamental tension between standardization and precision treatment is not a problem limited to educational policy, though parents are often the most vocal. The controversy is common in medicine. For example, in opioid use treatment, “standard treatment” such as medication-assisted therapy, is still not disseminated widely. Why? Large swaths of the patient population require resources to ensure appropriate follow-up and adherence to medication, and likely more intensive psychotherapy.

In a typical treatment trial for opioids, such as the NIDA funded START study or POATS study, approximately half of the patients in treatment achieve abstinence, whereas the other half drop out. This is compared to about a 90 percent drop-out rate without the use of medication, making a compelling case that medication should be a standard treatment. However, what about the half of the patients who are already assigned medication and still failed? This sounds like the failure rate of an inner-city public school. They need more. We don’t know what that more is yet but explicit behavioral modification treatment, including contingency management similar to the Moskowitz strategy, is a possible pathway.

I plotted this out: approximately half of individuals receive treatment and behave in an expected linear way. A little more resource can overcome their existing severity of illness or disadvantage at baseline. About another half require exponential resources and much more carefully planned treatment. But if we think logically about it, we can probably figure something out.

Finally, a small number of people reside in the region that requires specialized treatment. Their treatment response is nonlinear; if you assign them to either linear or exponential region treatment, they will likely fail and may do worse than if you did nothing at all. The irony is, the specialized treatment may not necessarily require more resources, it just requires more creative solutions and precision medicine approaches.

Policy debates right now are woefully lacking in precision and vocally talk pass each other. The left focuses on the linear region, where proportional governmental compensatory actions are likely to work; the right focuses on the exponential region, where self-directed spending that tends to grow exponentially might, in some cases, be more efficacious.Meanwhile, linear governmental interventions will likely and have already in many cases, failed. Nobody cares about the “region of specialized treatment,” because that region represents a small complicated minority that wields no political power and would require specialized advocacy efforts. 

You are reading

Hooked on Patterns

Harlem Success Academy and Substance Abuse Treatment

The simple case for more precision in policy.

Being Bayesian in Insane Places

A brief discussion of the mathematics of Rosenhan experiments and follow-ups

Online Therapy: How Does This Work?

A review of this rapidly moving, possibly dystopic technology